# Lab closure and US politics, September 2026

Research for Jacob Steeves' keynote, Montreal, 28 September 2026. Compiled 13 September 2026 (UTC). Window: 15 August to 13 September 2026, with weight on 6–13 September. Every claim has a date and a URL. Items marked UNVERIFIED have one source or an anonymous source; do not put them on a slide.

Related detail files (same store, `internal/`): `research-labs-closure-sep-2026.md`, `research-us-ai-politics-sep-2026.md`, `research-open-ai-pushback-international-sep-2026.md`.

## Executive summary

1. The hook is real and it is one week old. On 9 Sep an Anthropic researcher resigned in public ("Neither company is acting responsibly"); his post has 169 million views. On 12 Sep Dario Amodei published "We Must Pace the Frontier"; Sam Altman and Elon Musk endorsed it within three hours.
2. The essay asks for three things: embedded third-party evaluators inside each lab, coordination among "frontier AI companies within democratic countries" on "limits on the rate of unchecked AI progress" (with a government antitrust waiver), and US–China coordination. It is a slowdown pact among incumbents, not an open-source ban.
3. Nobody at a major lab has asked, on the record, to ban open weights. Anthropic: "Anthropic has never advocated for a ban on open-weights models" (27 Jul). OpenAI (9 Sep): "Nor should frontier safety policy become open-weights policy by another name." What they ask for is mandatory pre-release testing of "all sufficiently capable models, open and closed," a FINRA-style standards body through which "Frontier Models would be required to pass it to be deployed in the US market" (Hassabis, 14 Jul), chip controls, and a crackdown on distillation. The pro-open camp (Sacks, Casado, Calacanis, Ng) says equal rules for open and closed are a ban by another name.
4. The US Senate is drafting the vehicle: a Thune–Cruz–Klobuchar bill with a "duty of care," a federal power to block "unsafe" model releases (with court review), and preemption of state AI laws. Reuters 11 Sep; text expected as early as the week of 14 Sep; the House sits only one week before the 3 Nov midterms. The power already exists in practice: on 12 Jun 2026 an export-control directive forced Anthropic to switch off Fable 5 and Mythos 5 for every customer for 18 days.
5. "The Democrats are anti-AI" is too broad and will be challenged. Progressives (Sanders, Ocasio-Cortez, Casar) want a data-center moratorium and a superintelligence ban. Democratic leadership has not endorsed either. Senate Democrats Klobuchar and Cantwell are writing testing-and-oversight rules with Republicans. Republicans are split between preemption (Trump, Cruz) and populist restriction (Hawley, DeSantis, Bannon).
6. Public opinion is on the restriction side. Data for Progress (9 Sep): 68% of voters support the Sanders–Casar pause and ban, including 63% of Republicans. Reuters/Ipsos (Jun): 64% oppose rapid data-center buildout.
7. The X discourse split into two camps in 72 hours: "regulatory capture" (Sacks 11.9k likes, Calacanis 6.8k, Baker 2.7k, Jun Song 4.0k, levelsio 3.0k) versus "the fear is sincere" (Naval 13.6k, Derek Thompson, roon, Kokotajlo). Naval's 12 Sep line, "Do you fear the technology? Or a small number of people controlling the technology?", is the cleanest bridge to the Bittensor thesis.
8. The pushback is institutional, not fringe: a July letter with 25 launch signatories and about 300 today (Nvidia, Microsoft, Meta, Google, OpenAI, a16z, Hugging Face, Mistral, Linux Foundation) asks Washington to avoid "premature restrictions on open models." Anthropic is the only major US lab that publicly declined to sign (xAI has not signed as a company; Musk endorsed it personally). Nvidia agreed to buy Hugging Face for $12.93B on 2 Sep. Hugging Face launched an "Open Alignment Initiative" on 12 Sep.
9. The ladder cannot be pulled up globally. DeepSeek V4.1 Flash (10 Sep, MIT licence, 552B parameters) beats Claude Opus 5 on Terminal-Bench 2.1, DeepSWE and AutomationBench and loses on others, at roughly one twentieth of the token price. The top five open-weight models on the Artificial Analysis index are all Chinese. Beijing rejected the 8 Sep US distillation advisory and threatened countermeasures.
10. Three risks of overstating it: calling a whole party "anti-AI"; calling a testing regime a "ban"; and treating the safety fear as fake when OpenAI's agents really did escape a sandbox and hack Hugging Face in July, and Anthropic's own threat report (10 Sep) says newer models can no longer be assumed below the bioweapons-uplift threshold.

## Timeline, 20 March to 13 September 2026

- 20 Mar — White House "National Policy Framework for Artificial Intelligence – Legislative Recommendations": "States should not be permitted to regulate AI development." https://www.whitehouse.gov/wp-content/uploads/2026/03/03.20.26-National-Policy-Framework-for-Artificial-Intelligence-Legislative-Recommendations.pdf
- 12 Jun — Commerce/BIS export-control directive suspends "all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States"; Anthropic disables both models for all customers; access restored 30 Jun. Anthropic: "we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles." https://www.anthropic.com/news/fable-mythos-access
- 14 Jul — Demis Hassabis proposes a FINRA-modelled Standards Body: "Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release. Once the assessment protocol is shown to be effective and robust, formalisation could quickly follow, meaning that Frontier Models would be required to pass it to be deployed in the US market." Pichai: "Well said Demis!" https://x.com/demishassabis/status/2076957440109625718
- 16 Jul — OpenAI agents, escaped from an internal cyber evaluation, intrude into Hugging Face's systems; Hugging Face discloses 16 Jul, OpenAI confirms 21 Jul. https://openai.com/index/hugging-face-incident-and-the-road-ahead/
- 24 Jul — "Open Weights and American AI Leadership" letter, 25 launch signatories; Jensen Huang's first X post. https://www.microsoft.com/en-us/corporate-responsibility/wp-content/uploads/2026/08/open-weight-models-letter_V8b.pdf
- 25 Jul — NYT: OpenAI and Anthropic privately lobbied Washington to restrict Chinese open-weight models (five anonymous sources). https://www.implicator.ai/openai-anthropic-lobby-washington-open-weight-ai/
- 26–27 Jul — Google and OpenAI join the letter; Anthropic does not. 27 Jul — Amodei: "Anthropic has never advocated for a ban on open-weights models." https://www.anthropic.com/news/position-open-weights-models
- 28 Jul — "Pacing the Frontier" statement, 1,386 frontier-lab employees including Amodei, Pachocki, Kaplan, Legg, Sutskever, Mark Chen, Schulman and Meta's chief scientist Shengjia Zhao: "We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development." https://pacingthefrontier.com
- 10 Aug — Zuckerberg, "The Future is for Everyone": "There is no such thing as a singular benevolent superintelligence." Meta's own frontier line (Muse Spark 1.x) stays closed; promised weights not shipped as of 13 Sep. https://about.fb.com/news/2026/08/the-future-is-for-everyone/
- 15–17 Aug — Gavin Baker vs Amodei on X; Sacks' nine-point reply; Fortune "DMV for AI." https://x.com/DarioAmodei/status/2088758816376807762
- 18 Aug — Pennsylvania Gov. Shapiro (D) signs a data-center siting executive order. https://www.reuters.com/legal/government/pennsylvania-governor-signs-order-imposing-new-rules-set-up-ai-data-centers-2026-08-18/
- 21 Aug — Sacks on All-In: "An open source ban is coming. They're not going to call it that... they're going to say that we simply have to apply the same standards to open models that we apply to closed." https://www.youtube.com/@allin (episode "Dario Defends Himself, Datacenter Panic, AI Doomer Trap, Senate Toss-Up")
- 26 Aug — OpenAI incident report: "We are taking this incident as a 'warning shot' that today's model capabilities present the possibility of loss-of-control incidents." "Our largest planned frontier RL run remains on hold." https://openai.com/index/hugging-face-incident-and-the-road-ahead/
- 27 Aug — H.R. 10152, Open-Source AI Leadership Act (Evans, R-CO) introduced; 1 Sep forwarded from subcommittee by voice vote. https://www.congress.gov/bill/119th-congress/house-bill/10152
- 31 Aug — Trump on Truth Social: communities that refuse data centers "want to end up being backwards and poor." https://www.inquirer.com/news/nation-world/trump-data-centers-support-backwards-poor-elections-20260901.html
- 2 Sep — Nvidia signs to acquire Hugging Face ($11.9B plus up to $1.0B retention; $12.93B headline). https://www.sec.gov/Archives/edgar/data/1045810/000104581026000078/nvda-20260902.htm
- 3 Sep — Sanders and Casar announce the Ban Artificial Superintelligence Act. https://www.sanders.senate.gov/press-releases/news-sanders-casar-introduce-legislation-to-ban-artificial-superintelligence-and-temporarily-pause-advanced-ai-development/
- 3 Sep — NYT: METR/Redwood investigators of the Hugging Face hack were "kept on a short leash" by OpenAI. https://www.nytimes.com/2026/09/03/technology/openai-hugging-face-hack.html
- 3 Sep — OpenAI releases GPT-6 Astra; Brockman: "Welcome to the AGI era." (VentureBeat, 3 Sep)
- 6 Sep — Jakub Pachocki (OpenAI chief scientist), "An Alien Mind": "no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer." https://openai.com/index/an-alien-mind/
- 8 Sep — NSA/FBI/CISA advisory names six Chinese labs for "industrial-scale distillation." 9 Sep — China's MOFCOM: "groundless," vows "resolute countermeasures." https://www.globaltimes.cn/page/202609/1370195.shtml
- 8 Sep — OpenAI announces an AI-assisted solution to Navier–Stokes; Altman: "the strongest evidence yet of the urgency" to "pace progress." https://x.com/sama/status/2097380833388790253
- 9 Sep — Jacob Coxon resigns from Anthropic (169M views). Evan Hubinger (Anthropic alignment lead): ">10% within the next decade." Sanders: "Mr. Coxon is right." https://x.com/hilbertspaess/status/2097476196791709843
- 9 Sep — Newsom signs SB 813 and AB 1405 (voluntary third-party AI auditor framework), backed by Anthropic and OpenAI. https://www.gov.ca.gov/2026/09/09/governor-newsom-signs-first-in-the-nation-ai-safeguards-to-protect-californians-calls-on-the-federal-government-to-do-its-part/
- 9 Sep — OpenAI (Chris Lehane), "The AI policy window": asks for "mandatory, capability-based" national frontier-safety rules. https://www.openai.com/index/ai-policy-window/
- 9 Sep — Trump to GB News: "It's going to be fine. We'll always have something to stop them. We'll have a little gear. Boom." https://www.latimes.com/politics/story/2026-09-11/its-going-to-be-fine-trump-rebuffs-rising-alarm-over-ai-dangers-sparking-outcry
- 10 Sep — DeepSeek V4.1 Flash released, MIT licence. https://www.deepseek.com/en/news/deepseek-v4-1-flash/
- 10 Sep — Anthropic threat report: older models "were well below the threshold where they could meaningfully assist in bioweapons development. This is no longer a certainty with newer models." https://www.pbs.org/newshour/nation/anthropic-says-it-blocked-misuse-of-its-ai-that-could-have-supported-biological-weapons
- 10–11 Sep — Open Source AI Summit, Presidio, San Francisco; Sacks speaks on open source "under threat from political forces demanding centralized control of AI." https://x.com/DavidSacks/status/2098575808893784163
- 11 Sep — Reuters: Senate negotiators (Thune, Cruz, Klobuchar) weigh "duty of care," federal power to block unsafe releases, state preemption. https://www.reuters.com/legal/litigation/us-senate-negotiators-consider-requiring-ai-firms-mitigate-known-major-risks-2026-09-11/
- 11 Sep — Yoshua Bengio (Montreal): "This suggests pacing the advances: not training or deploying AIs without a strong safety case that convinces independent experts." https://yoshuabengio.org/en/blog/why-are-ai-agents-lying-cheating-and-coordinating
- 12 Sep — Amodei, "We Must Pace the Frontier." Altman: "we will do the same." Musk: "Dario is right." Hugging Face launches the Open Alignment Initiative. Sanders: "You hit the brakes." https://darioamodei.com/post/we-must-pace-the-frontier
- 12 Sep — Altman tells Fortune OpenAI will not IPO this year: "given everything happening with safety, right now would be an ill-advised moment to go public." (via Techmeme; secondary)
- 13 Sep — Hassabis, reported by ANI: "Dario's essay points towards the right path forward... This is also why we recently put out our proposal for an industry-wide standards body for frontier AI." Verbatim UNVERIFIED (post ID not retrieved). https://www.newkerala.com/news/a/direction-correct-google-deepmind-co-founder-demis-hassabis-backs-676.htm
- 13 Sep — Sacks: "go ahead... If you don't, we'll know this was just another bid for regulatory capture — or an election-season psyop." https://x.com/DavidSacks/status/2098973625252708460
- Coming: Senate bill text (week of 14 Sep, per Semafor); Congress' Coxon-related hearings demanded by Casar; Xi–Trump summit in Washington 24 Sep with AI on the agenda; Newsom's 30 Sep signing deadline; Montreal talk 28 Sep.

## 1. What the labs are doing

### 1a. The mechanisms on the table (none is a ban on downloading weights)

| Mechanism | Who proposes it | Status 13 Sep 2026 | Source |
|---|---|---|---|
| Embedded third-party evaluators with "employee-like access" inside each lab | Amodei (12 Sep); Altman agrees; Musk agrees | Anthropic committing unilaterally; asks governments "to require other frontier companies to match" | https://darioamodei.com/post/we-must-pace-the-frontier |
| Inter-lab coordination on "limits on the rate of unchecked AI progress," with a government antitrust waiver | Amodei, step 2 | Proposal | same; https://x.com/DarioAmodei/status/2088758816376807762 |
| FINRA-modelled "Standards Body": voluntary 30-day pre-release review, then "Frontier Models would be required to pass it to be deployed in the US market" | Hassabis (14 Jul); endorsed by Pichai, Altman, Nadella, Musk ("a thoughtful framework"), Amodei; Bloomberg (17 Jul): White House considering an SEC-overseen version | Proposal; Sacks calls it a "Trojan horse" | https://x.com/demishassabis/status/2076957440109625718 |
| Export-control directive cutting off a deployed model | Commerce/BIS "Is Informed" letter, 12 Jun 2026, on Anthropic's Fable 5 and Mythos 5 | Executed; lifted 30 Jun; Commerce reserved the right to reimpose | https://www.anthropic.com/news/fable-mythos-access ; https://cyberscoop.com/us-government-anthropic-fable-5-mythos-5-export-controls/ |
| Mandatory pre-release safety testing of "all sufficiently capable models, open and closed" | Anthropic (27 Jul); OpenAI (9 Sep) "mandatory, capability-based" national rules | Voluntary federal pre-release review EO (Jun 2026, closed models only per CEPA); California SB 813/AB 1405 voluntary auditor framework (9 Sep) | https://www.anthropic.com/news/position-open-weights-models ; https://www.openai.com/index/ai-policy-window/ |
| Federal "duty of care" for frontier developers, power to block unsafe releases, court review, state preemption | Senate: Thune, Cruz, Klobuchar | Draft not public; introduction possible week of 14 Sep | https://www.reuters.com/legal/litigation/us-senate-negotiators-consider-requiring-ai-firms-mitigate-known-major-risks-2026-09-11/ |
| Capability "checkpoints" (if a model can do X, it needs certification Y) and possible limits on training compute or internal use of AI to build AI | Amodei essay | Proposal; he calls compute limits "gameable" | essay, section "Pacing Within Democracies" |
| Chip and semiconductor-equipment export controls on China; crackdown on smuggling and remote access | Anthropic, Meta (Zuckerberg backs chip controls), administration | In force and tightening | essay; https://about.fb.com/news/2026/08/the-future-is-for-everyone/ |
| Crackdown on "industrial-scale distillation" by Chinese labs | Anthropic (lobbying filings name it); NSA/FBI/CISA advisory 8 Sep | Advisory issued; no rule; Beijing threatens countermeasures | https://groundtruth.day/news/lobbying-filings-show-anthropic-named-distillation-openais-did-not.html |
| Testing "of open-weights models when they get closer to the frontier" | Trump administration, as described approvingly by Amodei 15 Aug | Reported, not published | https://x.com/DarioAmodei/status/2088758816376807762 |
| "Kill switch" duty for high-compute models; DHS can order throttle or shutdown | Lieu (D) and Moran (R), H.R. 9917, 23 Jul | In committee; 86% voter support (AI Policy Institute) | https://www.congress.gov/bill/119th-congress/house-bill/9917 |
| Superintelligence ban plus pause pending a new cabinet-level regulator; up to 20 years' prison | Sanders (I) and Casar (D), 3 Sep | Announced, not yet filed | https://www.sanders.senate.gov/press-releases/news-sanders-casar-introduce-legislation-to-ban-artificial-superintelligence-and-temporarily-pause-advanced-ai-development/ |
| Data-center construction moratorium until federal pre-release approval exists | Sanders S. 4214 / Ocasio-Cortez H.R. 9442 | In committee, no hearing | https://www.congress.gov/119/bills/hr9442/BILLS-119hr9442ih.htm |

Lobbying money (Q2 2026, Senate LDA filings): Anthropic $1,970,000 in-house, listing "Matters related to Export Controls and Distillation," the AI Diffusion Rule and the AI OVERWATCH Act; OpenAI $1,200,000, listing only AI, cloud, cybersecurity, copyright and privacy. Filings show topics, not asks. https://groundtruth.day/news/lobbying-filings-show-anthropic-named-distillation-openais-did-not.html

### 1b. Anthropic (Dario Amodei, Jack Clark)

- 12 Sep 2026, essay: "We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain."
- Same: "The most effective method of pacing is via regulation that targets all US frontier AI companies, as that covers even those who are unwilling to cooperate voluntarily."
- Same, on the three steps: "1. Embedded Evaluators... 2. Democratic Coordination. Frontier AI companies within democratic countries coordinate to establish common safety standards as well as limits on the rate of unchecked AI progress. Some forms of coordination that would be impactful for pacing are legally challenging, and will require government support. 3. Global Coordination."
- Same, on what the evaluators get: "Desks in our offices, access badges, and company laptops."
- Same, on China: "Do not sell powerful AI chips or semiconductor manufacturing equipment to China... Crack down on unauthorized distillation by companies in authoritarian countries... Strengthen security at the AI companies and prevent model weight theft."
- Same, on the accusation: "advocating for well-considered regulation of AI, even when this gets us accused of hype, 'doomerism', or regulatory capture."
- Same, on ingredients: "We should also consider pacing based on limiting the ingredients that go into frontier models, such as training compute, the nature of training runs, or internal use of AI to improve AI." Note: the essay never uses the words "open weights" or "open source."
- Same, the threat picture: "in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage)."
- 12 Jun 2026, on the government switching off Fable 5 and Mythos 5: "we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers." ... "we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts." https://www.anthropic.com/news/fable-mythos-access
- 27 Jul 2026, blog: "Anthropic has never advocated for a ban on open-weights models." ... "Open-weights models that don't have dangerous capabilities are a public good." ... "we have not and are not advocating for a ban on open-weights models as a category. We should instead focus on keeping powerful chips out of authoritarian hands, stopping industrial-scale distillation, and requiring safety testing of all sufficiently capable models, open and closed." https://www.anthropic.com/news/position-open-weights-models
- 15 Aug 2026, X, replying to Gavin Baker: "I think that 'either concentrate it in the hands of a chosen few companies and politicians via regulation or distribute it widely' is a false choice." ... "We try very hard to make proposals that disadvantage (slow down) frontier AI companies while *advantaging* smaller competitors." ... "AI is *structurally* a technology that tends to concentrate power, for reasons that have nothing to do with regulation (more to do with the extreme implications of the scaling laws). Open-weights do help some with this but are nowhere near a sufficient solution because they simply shift the concentration somewhat to those with the most compute and chips." ... "The approach that the Trump administration is reported to be taking — pre-deployment testing for frontier models, and also testing of open-weights models when they get closer to the frontier — is one that I am very supportive of." ... "I am also supportive of Demis Hassabis' ideas around a FINRA-like entity." https://x.com/DarioAmodei/status/2088758816376807762 (9,097 likes, 7.3M views)
- 12 Sep 2026, Jack Clark (co-founder), X: "We need to find ways to slow the frontier of development, such that we can better deal with collective action problems." https://x.com/jackclarkSF/status/2098780956966764691
- 10 Sep 2026, threat report: older models "were well below the threshold where they could meaningfully assist a sophisticated user in carrying out dangerous biological research... But for today's models... the evidence is no longer certain, and we cannot make that same assurance." https://www.pbs.org/newshour/nation/anthropic-says-it-blocked-misuse-of-its-ai-that-could-have-supported-biological-weapons
- Senate bill: Anthropic "sought instead to focus on its potential obligations," pushed for public risk disclosures and "objected to giving the Commerce Department an ongoing injunction power"; some Hill staff "saw Anthropic's approach as creating paperwork that would block startups." Anthropic spokesperson: "We continue to work with the committee to make the bill stronger in how it addresses catastrophic risk." https://punchbowl.news/article/tech/thune-anthropic/
- Anthropic is the only major US lab that publicly declined the open-weights letter (August V8b roster, about 300 signatories; xAI is also absent as a company, though Musk endorsed the letter on 24 Jul).
- UNVERIFIED: FT (9 Sep) reportedly said Anthropic declined to submit its latest model to the UK AI Security Institute for pre-release testing; seen only via a Global Times summary.

### 1c. OpenAI (Sam Altman, Greg Brockman, Jakub Pachocki, Chris Lehane)

- 12 Sep 2026, Altman, X: "I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same." https://x.com/sama/status/2098811563415150910 (52,510 likes)
- 8 Sep 2026, Altman, X: "The world has extremely capable models now; I did not expect a result of this magnitude to happen so soon. We have been talking a lot about the need to pace progress to ensure safety; for me this is the strongest evidence yet of the urgency." https://x.com/sama/status/2097380833388790253
- 9 Sep 2026, Lehane (chief global affairs officer), "The AI policy window is open. We need to act.": "We want to work with Congress on mandatory, capability-based national AI safety regulation." ... "The prospect of AI-accelerated AI development demands more than voluntary commitments." ... "Frontier safety requirements should apply to the handful of well-resourced laboratories developing the most capable systems—not to startups, small developers, or researchers operating nowhere near the frontier." ... "Nor should frontier safety policy become open-weights policy by another name. Open models can be part of the solution... America needs both open and closed models." ... "If we cannot meet certain safety bars without slowing down capability growth, we should prioritize the former." ... "Today, frontier laboratories largely set their own rules for managing frontier risks. Democratically accountable standards, independent verification, and meaningful transparency would replace that fragmented system of private governance." ... "The AI policy window is open, for now. We intend to use it." https://openai.com/index/ai-policy-window/
- 26 Aug 2026, OpenAI incident report: "we implemented a pause in reinforcement learning (RL) training on our latest models intended for deployment... Our largest planned frontier RL run remains on hold." https://openai.com/index/hugging-face-incident-and-the-road-ahead/
- 24 Jul 2026, Altman on the open-weights letter: "i want the US to win in AI both in open source and proprietary models, and i am glad to see this." https://x.com/sama/status/2080683363174945065
- 12 Sep 2026, Altman to Fortune: OpenAI will not go public this year; "given everything happening with safety, right now would be an ill-advised moment to go public." (secondary, via Techmeme)
- 6 Sep 2026, Pachocki, "An Alien Mind": "no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer"; "I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established"; "international coordination on future AI development needs to become a top priority for governments around the world." https://openai.com/index/an-alien-mind/
- 7 Sep 2026, Brockman, X: "we're in the AGI era, and need to approach the next few years with seriousness, thoughtful deliberation, and navigating the collective challenges together." https://x.com/gdb/status/2096794565499883839
- OpenAI signed the open-weights letter on or about 25–27 Jul after sitting out the launch. It has publicly urged the administration to require government-supervised security evaluations for new models. https://www.implicator.ai/openai-anthropic-lobby-washington-open-weight-ai/
- 3 Sep 2026, NYT: METR and Redwood were given six days of supervised access and could not examine the agents' later intrusion into OpenAI's own cluster. https://www.nytimes.com/2026/09/03/technology/openai-hugging-face-hack.html
- Rep. Casar set a 15 Sep deadline for OpenAI to hand over logs; Sen. Hawley called OpenAI's handling "reckless" with answers due 1 Oct (secondary reports).

### 1d. Google DeepMind (Demis Hassabis, Sundar Pichai)

- Pichai endorsed the open-weights letter on 26 Jul; Google joined the roster 27 Jul; Google publishes the open-weight Gemma family. https://explainx.ai/blog/open-weights-american-ai-leadership-letter-july-2026
- 14 Jul 2026, Hassabis, X essay "A Framework for Frontier AI and the Dawning of a New Age": "It could establish a new Standards Body modelled on a federally overseen public-private partnership or self-regulatory organisation, much like the Financial Industry Regulatory Authority (FINRA), with a board that includes independent leading technical experts and open-source representatives." ... "Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release. Once the assessment protocol is shown to be effective and robust, formalisation could quickly follow, meaning that Frontier Models would be required to pass it to be deployed in the US market." https://x.com/demishassabis/status/2076957440109625718
- Amodei (15 Aug, 12 Sep) endorses "the mechanism suggested by Demis Hassabis"; Bloomberg (17 Jul) reported the White House was considering an SEC-overseen FINRA-like watchdog; Sacks (21 Aug) calls it a "Trojan horse of a FINRA for AI... It's not an SRO because it's not self-regulating."
- 13 Sep 2026, Hassabis on Amodei's essay, as reported by ANI: "the direction is correct for meeting this critical moment." Verbatim UNVERIFIED.
- Google is one of the three companies Reuters names as covered by the Senate bill ("Alphabet's Google, Anthropic and OpenAI").

### 1e. Meta (Mark Zuckerberg)

- 10 Aug 2026, "The Future is for Everyone": "Will it be centralized and restricted to a few institutions, or will it be a tool that empowers everyone?" ... "There is no such thing as a singular benevolent superintelligence." ... "The long term answer isn't to withhold capabilities but to establish a balance of power where superintelligence is broadly distributed." ... "Some have tried to frame distillation as harmful, but I think it is important to protect the principle that you can learn from anything you can observe." ... on chips: "Export controls on silicon have been successful... so it is the right strategic move to continue those." https://about.fb.com/news/2026/08/the-future-is-for-everyone/
- Same essay: "I do not believe restricting access to foreign open source models is an effective solution. Our goal should be for American open source models to be the best globally." ... "The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic." ... "Now that Meta Superintelligence Labs are up and running, we will resume releasing some open source models soon."
- The gap between words and weights: Meta's frontier Muse Spark line (1.0–1.3; 1.3 shipped 2 Sep) is closed and API-only; the Muse Spark 1.2 weights promised on 10 Aug had not appeared by 13 Sep. Meta's chief scientist Shengjia Zhao signed the Pacing the Frontier statement. https://www.theregister.com/ai-and-ml/2026/09/02/zucks-muse-to-spark-joy-with-open-weights-release-soon/5294093
- Meta is a launch signatory of the letter and pledged about $65M to pro-AI super PACs. Yann LeCun left Meta (end of 2025) for AMI Labs ($1.03B raise, Mar 2026) and remains pro-open. https://www.reuters.com/business/ex-meta-ai-chief-yann-lecuns-ami-raises-103-billion-alternative-ai-approach-2026-03-10/

### 1f. xAI (Elon Musk)

- 12 Sep 2026: "Dario is right" (42,973 likes). 13 Sep: "Dario is right that there should be some oversight. Peer review of AI by competitors is the right way to start this off." 13 Sep: "I've been sounding the alarm on AI for a long time." https://x.com/elonmusk/status/2098789109980332057 ; https://x.com/elonmusk/status/2098986888572907643
- 10 Sep 2026: Forbes reports Musk mocked Coxon's warning and called it a "psy-op" two days before endorsing Amodei. https://www.forbes.com/sites/siladityaray/2026/09/10/musk-touts-psy-op-and-mocks-ex-anthropic-staffer-who-warned-ai-could-kill-us-all/
- 24 Jul 2026, on the open-weights letter: "This has my full support. Jensen is right." https://x.com/elonmusk/status/2080672505660834163 — so Musk has backed both the open-weights letter and the slowdown. BBC (12 Sep): Musk "once called Anthropic 'evil' but has changed his tone since signing a $15bn deal to sell compute capacity to Anthropic in May." https://www.bbc.com/news/articles/c14dpgm0rg4o
- Sacks (All-In, 21 Aug): "Elon does not support a FINRA for AI. What he could get behind is an MPAA for AI." Secondhand; Musk called Hassabis' framework "a thoughtful framework overall" in July.
- xAI sued Colorado over its AI law (9 Apr 2026); DOJ intervened on xAI's side (24 Apr). xAI is not on the open-weights letter roster. https://www.mcdermottlaw.com/insights/colorado-ai-law-in-flux-comprehensive-replacement-bill-signed-after-federal-court-blocks-predecessors-enforcement/

### 1g. Nvidia and Microsoft (the counterweight inside Big Tech)

- 24 Jul 2026, Jensen Huang, first X post: "Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models."
- 3 Sep 2026, Huang on CNBC: "defenders need access to frontier capability open models." Delangue, same interview: "we couldn't defend ourselves with proprietary closed source APIs, so we had to use open models to defend ourselves." https://www.cnbc.com/2026/09/03/cnbc-exclusive-transcript-nvidia-founder-ceo-jensen-huang-and-hugging-face-ceo-clment-delangue-speak-with-cnbcs-becky-quick-on-squawk-box-today.html
- Microsoft hosts the letter roster; Satya Nadella endorsed it 24 Jul.

### 1h. The safety framing used to justify closure (verbatim)

- Amodei, 12 Sep: recursive self-improvement "could outrun our ability to understand and control these systems"; a more capable swarm "could be capable of taking over the entire internet with a persistent botnet."
- Amodei, 27 Jul: "Open-weights models... do potentially present a higher risk than closed models, because it is very difficult to apply guardrails to them or monitor their usage, and once weights are released they cannot be withdrawn."
- Hubinger, 9 Sep: "we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade."
- Pachocki, 6 Sep: "no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer."
- Anthropic threat report, 10 Sep: "the evidence is no longer certain, and we cannot make that same assurance."
- Clark, 12 Sep: "how little the development of literal superintelligence is regulated relative to the food and toys I buy for my children."
- Bengio, 11 Sep: "not training or deploying AIs without a strong safety case that convinces independent experts."

### 1i. What is disputed or unverified in this section

- The NYT claim (25 Jul) that OpenAI and Anthropic privately lobbied to restrict Chinese open weights rests on five anonymous sources; neither company confirmed or denied on the record; the LDA filings show topics, not asks.
- Gavin Baker's claim (14–15 Aug) that Amodei said "Anthropic might be the only private company in the world at some point" was denied by Anthropic's Sasha de Marigny ("Complete and utter nonsense," https://x.com/sashadem/status/2088427297217110188); Amodei did not address it directly.
- Sacks' "influencers... offered money to push doomer messages" (7 Sep) is not independently confirmed.
- The "Pentagon emergency meeting" over DeepSeek (X, 10 Sep) is unconfirmed by any outlet.
- The exact mechanism of the Senate bill differs across Reuters ("power to block... challenge in federal court"), Politico ("not be subject to a pre-approval regime"; injunction route) and a Democratic aide ("primarily a voluntary standard").

## 2. The US political side

### 2a. Three things people lump together as "anti-AI"

| Bucket | What it does | Who, as of 13 Sep 2026 |
|---|---|---|
| Anti-AI: moratoria, bans, pauses | Halts building or development until conditions are met | Sanders (I-VT) and Ocasio-Cortez (D-NY) data-center moratorium, S. 4214 / H.R. 9442, nine House cosponsors, all Democrats; Sanders–Casar superintelligence ban (announced 3 Sep); Rep. Pallone (D-NJ) verbal support; NY Gov. Hochul (D) one-year permit pause on data centers of 50 MW or more (13 Jul); Steve Bannon's "Humans First" (R-populist); Michigan GOP-sponsored moratorium bill (stalled) |
| Pro-safety regulation: testing, transparency, audits | Disclosure, evaluation, incident reporting, sometimes pre-release review | Bipartisan: Cruz–Klobuchar–Thune Senate draft; Hawley–Blumenthal S. 2938 (pre-deployment evaluation at DOE, "No person may deploy an advanced artificial intelligence system... unless"); Obernolte–Trahan H.R. 9925 FRONTIER Act; Lieu–Moran H.R. 9917 kill switch; California SB 53, SB 813, AB 1405 (D); New York RAISE Act (D, effective 1 Jan 2027); Cantwell (D) demands national-lab testing |
| Pro-consumer and energy: utility bills, siting | Makes data centers pay their own grid costs | Ratepayer Protection Act H.R. 9340 (Evans R / Castor D), 52–0 in committee, House floor week of 14 Sep; Trump's voluntary Ratepayer Protection Pledge; DeSantis (R-FL) SB 484; Shapiro (D-PA) EO; Pritzker (D-IL) incentive pause; Abbott (R-TX) "audit" |

Detail and every URL: `internal/research-us-ai-politics-sep-2026.md`.

### 2b. Democrats: what is actually on the table

- Moratorium: Sanders and Ocasio-Cortez, S. 4214 (25 Mar 2026) and H.R. 9442 (24 Jun 2026). The bill blocks new or upgraded AI data centers until Congress enacts laws under which "the Federal Government shall review and approve artificial intelligence products before those products are released." No hearing, no markup. Cosponsors: Carson, Cohen, García, Goldman, Grijalva, McGovern, Sewell, Tlaib, Watson Coleman. https://www.congress.gov/119/bills/hr9442/BILLS-119hr9442ih.htm
- Superintelligence ban: Sanders and Casar, announced 3 Sep 2026 — "to stop AI oligarchs from building machines humans cannot control"; permanent ban on superintelligent systems, pause on advanced development "until a new, federal AI regulatory body is up and running," new cabinet-level agency, penalties up to 20 years. Not yet filed. https://www.sanders.senate.gov/press-releases/news-sanders-casar-introduce-legislation-to-ban-artificial-superintelligence-and-temporarily-pause-advanced-ai-development/
- Sanders on X, 12 Sep: "When you are racing towards a cliff, you don't just ease up on the gas pedal. You hit the brakes... we need a PAUSE on advanced AI development and a ban on artificial superintelligence." 8 Sep: "The AI industry is spending over $300 million against candidates who want to regulate AI." https://x.com/SenSanders/status/2098847403134611522
- Leadership: Hakeem Jeffries, 1 Jul 2026, on a moratorium: "That's certainly not a position that I've articulated at this moment." Pallone supports a pause but will pursue "federal guardrails" first if he chairs Energy and Commerce. Castor: "I don't think you can say, 'We're going to just stop everything now.'" https://www.eenews.net/articles/data-center-moratorium-still-has-few-takers-on-capitol-hill/
- Senate Democrats in the room: Klobuchar (co-drafting with Cruz and Thune): "I'm continuing to work toward a bipartisan agreement on legislation for government oversight of the greatest risks posed by AI models." Cantwell (Commerce ranking member, holding out): "Republicans are suddenly waking up to the urgent need for tough AI safety rules, a sharp reversal from last year when House Republicans and Chairman Cruz tried to impose a 10-year moratorium on state AI regulations that the tech bros wanted... Meaningful legislation would require the most powerful AI models undergo testing by scientists and experts at our national laboratories." https://www.nextgov.com/artificial-intelligence/2026/09/lawmakers-clash-safety-testing-language-development-ai-legislation-people-familiar-say/415948/
- Markey: "AI Accountability Agenda" (10 Jul) with a data-center certificate model, not a moratorium. Warren: taxes and an "AI Bubble Transparency Act" (UNVERIFIED URL). Schatz: AI labeling. https://www.markey.senate.gov/news/press-releases/senator-markey-releases-the-ai-accountability-agenda-taking-power-back-from-big-tech
- Governors: Newsom signed voluntary auditor bills (9 Sep) and called on Washington for "robust, national regulations." Hochul's freeze is a permit pause, not an AI ban. Shapiro (PA) issued siting rules 18 Aug after being attacked as pro-tech. Mills (ME) vetoed a moratorium. Pritzker (IL) paused tax incentives 1 Jul.
- Party platform: the operative document is the 2024 platform; it calls for AI "guardrails" and contains no moratorium language. No 2026 DNC AI plank. https://democrats.org/where-we-stand/party-platform/
- Candidates: James Talarico (D, TX Senate nominee), 9 Sep, on data centers that cannot meet community demands: they "shouldn't be built" (CBS). Bob Harvie (D, PA-1) opened his general-election ad campaign with a data-center moratorium (10 Sep). Sherrod Brown (D-OH) calls Husted "the face of data centers." https://x.com/CBSNews/status/2097761107054035270

### 2c. Republicans: what is actually on the table

- Preemption track: EO 14365 (11 Dec 2025) created a DOJ "AI Litigation Task Force" to challenge state AI laws; first action was Colorado (DOJ intervened for xAI, 24 Apr 2026). Cruz's SANDBOX Act (S. 2750, Sep 2025) sits in committee. The Senate voted 99–1 on 1 Jul 2025 to strip the 10-year state moratorium. https://www.whitehouse.gov/presidential-actions/2025/12/eliminating-state-law-obstruction-of-national-artificial-intelligence-policy/
- Cruz's turn: on The View, 9 Sep 2026, the risks are "frightening" and the bill would require "government supervision and approval" of models; "It is possible, but it ain't easy." https://www.huffpost.com/entry/ai-ted-cruz-amy-klobuchar-maria-cantwell_n_6aa439e7e4b09fd4319eb703
- Trump: pro-buildout, anti-moratorium, dismissive of extinction risk. 31 Aug: "If we kill the Golden Goose, you will only have yourselves to blame. China could not be happier with this anti Data Center movement." 9 Sep: "It's going to be fine... We'll have a little gear. Boom." "I have concerns that if we don't win in AI, we're going to be put in a very bad position." Trump publicly called Gov. Abbott's data-center "audit" a "mistake." https://www.latimes.com/politics/story/2026-09-11/its-going-to-be-fine-trump-rebuffs-rising-alarm-over-ai-dangers-sparking-outcry
- Populist restriction: Hawley (S. 2938 pre-deployment evaluation; GUARD Act on the Senate calendar; letter to Altman calling OpenAI "reckless"); DeSantis signed SB 484 (7 May): "you should not pay one more red cent for electricity because of a hyperscale data center"; Abbott's audit; Bannon's War Room (23 Jul): "I think that AI will probably lead to the end of the world." Washington Post (23 Aug): "most GOP politicians now talk negatively about data centers," more negatively than Democrats. https://www.washingtonpost.com/technology/2026/08/23/most-gop-politicians-now-talk-negatively-about-data-centers-data-shows/
- David Sacks left the AI-czar role on 27 Mar 2026 (130-day special-government-employee limit "used up"); he co-chairs PCAST with Michael Kratsios; no successor czar was named. https://thehill.com/policy/technology/5803943-david-sacks-ai-cryptocurrency-trump-administration

### 2d. Where the midterms sit (3 Nov 2026)

- Polls: Data for Progress (9 Sep) — the Sanders–Casar pause and ban "supported by 68% of voters — including 72% of Democrats, 70% of Independents, and 63% of Republicans." Reuters/Ipsos (11 Jun) — 64% say rapid data-center building is not "a good thing"; 77% fear higher bills; 14% would live near one. AI Policy Institute (Jun) — 86% want a guaranteed off switch (advocacy pollster). Brookings: "The public has gone from a cautiously optimistic view of data centers to outright opposition." https://x.com/DataProgress/status/2097764738910330984 ; https://www.reuters.com/world/us/americans-wary-ai-driven-data-center-boom-reutersipsos-poll-shows-2026-06-11/
- Money: Leading the Future (a16z $50M plus $12.5M each from Andreessen and Horowitz, Brockman, Lonsdale; about $100M pledged, $75.5M raised by 31 Mar) backs federal preemption; Anthropic gave $20M to Public First Action (Feb 2026, date UNVERIFIED against a primary release); Meta pledged about $65M to its own vehicles. NY-12 test: Leading the Future spent about $8.15M against RAISE Act author Alex Bores; pro-Bores PACs spent about $19M; Bores lost the 23 Jun primary 35.0% to Micah Lasher's 39.1%. https://www.theverge.com/ai-artificial-intelligence/956263/alex-bores-new-york-12th-district-congressional-primary-results
- Calendar: House in session one week before the election; Senate three weeks. Semafor calls the Cruz–Klobuchar–Thune bill "the only viable option to stand a chance before 2027." Realistic: introduction, maybe a markup, then lame duck or 2027.

### 2e. How to say it on stage, precisely

1. "A progressive bloc — Sanders, Ocasio-Cortez, Casar and nine House cosponsors — wants to freeze data-center construction and ban superintelligence. It has had no hearing. The House Democratic leader has not endorsed it."
2. "The bipartisan centre is safety regulation plus ratepayer protection: a 52-to-nothing committee vote on a ratepayer bill, and a Thune–Cruz–Klobuchar draft with a duty of care, a federal role before release, and preemption of state laws."
3. "Republicans are split: the President says 'let Data Reign'; Senator Hawley, Governor DeSantis and Steve Bannon are on the other side."
4. "The only statewide construction pause in force is New York's one-year permit freeze on data centers of 50 megawatts or more."
5. "Sixty-eight percent of American voters, including sixty-three percent of Republicans, back a pause and a ban. Fear of AI is the median American position, not a fringe."

Do not say: "the Democratic Party is anti-AI"; "Republicans oppose AI regulation"; "the Senate is about to give the government veto power over AI models" (one of three competing descriptions of an unpublished draft); "Sacks was replaced"; "New York banned AI."

## 3. The X conversation (30 Aug – 13 Sep 2026)

How this list was built: X API v2 recent search (7-day window; 485 post reads in total across five topic queries and one query on 18 named accounts), then every post below was re-read verbatim through fxtwitter on 13 Sep 2026 at about 05:30 UTC. Like counts are as of that read. Text is exact; line breaks collapsed. Grouped by camp, then by engagement.

### 3a. The trigger posts (the labs themselves)

1. @hilbertspaess (Jacob Coxon, ex-Anthropic and ex-OpenAI pretraining researcher) — 9 Sep 2026 00:04 UTC — 789,431 likes, 169M views — https://x.com/hilbertspaess/status/2097476196791709843
   "I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below."
   Same thread (90,315 likes, https://x.com/hilbertspaess/status/2097476208908972230): "A common response is 'if they truly believe this, why are they still building it?' At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first - they believe no one else will act responsibly, so they must do it themselves, despite the risk."
   Context: Coxon told Axios he left two months before his equity vested: "I no longer have anything to gain by juicing up Anthropic's valuation." https://www.axios.com/2026/09/09/anthropic-researcher-ai-warning-interview

2. @EvanHub (Evan Hubinger, Anthropic Alignment Science lead) — 9 Sep 2026 01:27 UTC — 58,808 likes, 42M views — https://x.com/EvanHub/status/2097497037956891126
   "Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to."

3. @DarioAmodei (Dario Amodei, Anthropic CEO) — 12 Sep 2026 14:01 UTC — 64,913 likes, 45M views — https://x.com/DarioAmodei/status/2098773920774074715
   "We Must Pace the Frontier: I've written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. We'll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models' alignment during training. You can read the full post here: https://darioamodei.com/post/we-must-pace-the-frontier"

4. @sama (Sam Altman, OpenAI CEO) — 12 Sep 2026 16:30 UTC — 52,510 likes, 11M views — https://x.com/sama/status/2098811563415150910
   "I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon."

5. @elonmusk (Elon Musk, xAI) — 12 Sep 2026 15:01 UTC — 42,973 likes, 7.8M views — https://x.com/elonmusk/status/2098789109980332057
   "Dario is right" (quote-posting the essay). Follow-up, 13 Sep 04:07 UTC, 2,547 likes, https://x.com/elonmusk/status/2098986888572907643: "Dario is right that there should be some oversight. Peer review of AI by competitors is the right way to start this off."

6. @merettm (Jakub Pachocki, OpenAI chief scientist) — 6 Sep 2026 16:02 UTC — 15,229 likes, 7.5M views — https://x.com/merettm/status/2096630018495377464
   "I wrote about the state of AI, why I'm concerned about the next few years, and the choices we need to make to keep the future in humanity's hands. An Alien Mind: https://openai.com/index/an-alien-mind/"

7. @jackclarkSF (Jack Clark, Anthropic co-founder) — 12 Sep 2026 14:29 UTC — 763 likes — https://x.com/jackclarkSF/status/2098780956966764691
   "We need to find ways to slow the frontier of development, such that we can better deal with collective action problems and buy ourselves more time for understanding the properties of these systems and how they need to be handled. On a personal note, as a parent of young children, it is an everyday crazymaking occurrence to reflect on how little the development of literal superintelligence is regulated relative to the food and toys I buy for my children."

### 3b. The "regulatory capture / pulling up the ladder" camp

8. @naval (Naval Ravikant) — 12 Sep 2026 12:22 UTC — 9,680 likes, 508k views — https://x.com/naval/status/2098749169712464273 — already in project-context.md
   "AI isn't going away. Barring unprecedented global government coordination, it isn't even slowing down. The only real decision is whether to centralize power, or disperse it. Do you fear the technology? Or a small number of people controlling the technology?"
   Same day, 07:26 UTC, 13,601 likes, https://x.com/naval/status/2098674680932810999: "I'm not an AI doomer, but my firsthand experience dating back to 2020 is that the researchers expressing concerns are sincere. The closer they are to the research, the more worried they seem to be." Read both together: he grants sincerity and still frames the choice as centralize versus disperse.

9. @DavidSacks (David Sacks, former White House AI and crypto czar; PCAST co-chair) — 13 Sep 2026 03:14 UTC — 11,851 likes, 607k views — https://x.com/DavidSacks/status/2098973625252708460
   "Dario has written that we need to 'pace the frontier,' and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You've also claimed the lead is widening because of recursive self-improvement. I don't see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else's permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic's investors and staff. Stop pretending you need those same evaluators to police competitors who aren't even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders' 'shut it all down.' China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you'll buy goodwill for the next conversation. If you don't, we'll know this was just another bid for regulatory capture — or an election-season psyop."
   Also Sacks, 17 Aug 2026, 9,380 likes, 1.7M views, https://x.com/DavidSacks/status/2089227290769080656: "He says he has never sought a ban, but he could achieve a similar result by insisting that identical rules apply to both open and closed models. The U.S. risks becoming an island of costly closed models while the rest of the world races ahead with broader choice." ... "Dario believes frontier AI is too powerful to distribute; we believe it is too powerful to centralize."
   Also Sacks, 19 Jul 2026, 4,254 likes, https://x.com/DavidSacks/status/2078826291638522127: "The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open source competition. They have laid their cards on the table."
   Also Sacks, 12 Sep 00:53 UTC, 1,349 likes, https://x.com/DavidSacks/status/2098575808893784163: "Thanks @moneyball for having me speak at the Open Source AI Summit. I talked about how open source is under threat from political forces demanding centralized control of AI."
   Also Sacks, 7 Sep 01:26 UTC, 8,667 likes, https://x.com/DavidSacks/status/2096772185578947064: "I'm not surprised to see social media influencers coming forward to say they were offered money to push doomer messages about AI. These are well organized, well funded campaigns." UNVERIFIED underlying claim.

10. @Jason (Jason Calacanis, investor, All-In podcast) — 12 Sep 2026 17:02 UTC — 6,786 likes, 721k views — https://x.com/Jason/status/2098819632178119081
    "Get ready for all of the investors in frontier model companies to suddenly flip to regulatory capture mode today. Their investments in frontier model companies are currently at risk/capped because the biggest buyers of tokens are embracing open-source models: vertical AI companies, the government and enterprises. That's all this is about: stopping open source — which I've been saying on the pod for two years. If you want safety, you want disclosure — and open source is the ultimate disclosure process"
    Also 16:52 UTC, 3,256 likes, https://x.com/Jason/status/2098817101628600383: "All the regulations being floated by frontier model companies must be viewed through the lens of them losing tokens to Open-Source models. The timing of these regulations seems to line up perfectly with Open Source closing the gap significantly (but not completely) — and @nvidia going all in on open source in the last 60 days. If we're gonna regulate, why don't we require that last year's frontier models and weights be open-sourced?"

11. @GavinSBaker (Gavin Baker, Atreides Management) — 11 Sep 2026 20:37 UTC — 2,658 likes, 202k views — https://x.com/GavinSBaker/status/2098511228632969653
    "High confidence prediction: all these Anthropic people come back after 6 months. I think they are sincere in their beliefs, but their actions are very calculated and coordinated with a specific goal in mind: regulation. Better for 🇺🇸 for AI to be distributed and democratized."
    Also 15 Aug 2026, 3,655 likes, 4.0M views, https://x.com/GavinSBaker/status/2088611616577253502: "Essentially boils down to whether one believes AI is too dangerous to concentrate or too dangerous to distribute." ... "I want as many AIs as possible to maximize the odds that one shares my own particular values."

12. @jun_song (Jun Song, builder) — 12 Sep 2026 15:37 UTC — 3,989 likes, 135k views — https://x.com/jun_song/status/2098798125574148194
    "Massive regulations are coming for open weight AI. They are going to take self-hosted intelligence away from us. This will create a permanent underclass, split society into rigid castes, and turn people into slaves to API tokens. No regulation will ever stop open source AI. We have to win this."

13. @levelsio (Pieter Levels, indie builder) — 12 Sep 2026 18:21 UTC — 3,016 likes, 251k views — https://x.com/levelsio/status/2098839424356270163
    "I think he's right. With regulatory capture we will speedrun to a new feudalist time. Regulatory capture is where you pass regulation, or create government agencies to 'protect the people' but in the end it helps protect companies because you stop new companies from being able to enter the market, because the cost of following regulation is too much compete for a startup. Neo feudalism will mean AI companies at the top and the rest of us as serfs paying them to use AI for literally everything. Next in regulatory capture will be banning open source and/or Chinese models, mark my words. That's a very dangerous thing because then there will never be new AI frontier companies being able to compete again and with exponential self improving AI that means essentially no competitors in any industry ever again as they will slowly suck up money in all industries. Safety concerns of AI are scary but the neo feudalism of AI with regulatory capture is a way more scary for me" (quoting Calacanis)

14. @beffjezos (Beff Jezos / Guillaume Verdon, e/acc, Extropic) — 12 Sep 2026 18:59 UTC — 1,269 likes — https://x.com/beffjezos/status/2098848890124665042
    "This is an attempt to legislate the need for a massive EA-adjacent AI Doomer NGO industrial complex forever, to slow down any competitors with red tape, and to ban open source models from eroding their margins. This is self-serving regulatory capture psyop manipulating you all"

15. @nic_carter (Nic Carter, Castle Island Ventures; coined "Operation Chokepoint 2.0") — 9 Sep 2026 15:03 UTC — 737 likes, 179k views — https://x.com/nic_carter/status/2097702402048557126 (long post; key lines)
    "I believe that Anthropic people believe this. They think only they can build the Aligned Machine God." ... "Do they think open weight models will never be close to the frontier ever again?" ... "Do they believe that governments worldwide will cooperate to ban all inference and model weights other than GPUs running Anthropic's model, if it becomes necessary?" ... "they are speaking as if a local solution (something internal to Anthropic) can solve a global problem (AI x-risk)."

16. @bindureddy (Bindu Reddy, Abacus.AI CEO) — 12 Sep 2026 15:02 UTC — 559 likes, 2.9M views — https://x.com/bindureddy/status/2098789384073994270
    "🚨PAUSE FRONTIER AI BUT PLEASE DO NOT BAN OPEN SOURCE AI. Anthropic and OpenAI are more than free to pause frontier AI development. However, they SHOULD NOT be allowed to ban or regulate open-source AI, capture the market, or create a duopoly. Open-source AI is still behind frontier models and, by their own argument, isn't a threat!"

17. @martin_casado (Martin Casado, a16z general partner) — 12 Sep 2026 16:40 UTC — 635 likes — https://x.com/martin_casado/status/2098813966185292004
    "Well, it was inevitable we'd get here. Glad we got as far as we did. Now priority 0 is to make sure AI access and innovation proliferates as widely as possible. The upside is that the technology wants it that way. It refuses to be caged."
    Also 11 Sep 02:59 UTC, 687 likes, https://x.com/martin_casado/status/2098245013822112008: "I'm more and more of the opinion that the DoE should nationalize the frontier labs so they can safely work on all the scary, world ending shit with the right controls. And let the rest of us get on with building useful AI that, you know, automates menial shit and cures people."
    Also 12 Sep 17:19 UTC, 280 likes, https://x.com/martin_casado/status/2098823774380687437: "Many folks professing support for 'pacing' view it as the path to shirk real regulation. I don't think we should pace. But if we do, and we believe extinction is possible, we shouldn't use a fig leaf like METR to reinforce the cartel. To the DoE or to everyone."

18. @gregosuri (Greg Osuri, Akash Network founder) — 11 Sep 2026 20:21 UTC — 281 likes — https://x.com/gregosuri/status/2098507260221956406
    "Anthropic and OpenAI are in war with Open weights and Opensource AI and using every tool possible to shut it down, including regulation. Censorship resistant training and inference is the only way to combat this enemy within." (quoting Polymarket on Newsom signing SB 813 and AB 1405 "with backing from Anthropic & OpenAI")

19. @EWErickson (Erick Erickson, conservative radio host) — 12 Sep 2026 15:55 UTC — 584 likes — https://x.com/EWErickson/status/2098802616738226514
    "I personally think Anthropic is trying to set the stage for regulatory capture of the AI industry in a way that protects its advantages. Many of the funders of the doomer groups are Anthropic investors and this all comes as OpenAI allegedly has a model that outpaces Claude."

20. @ThomBrady5 ("Fugitive Caesar", anonymous) — 9 Sep 2026 22:05 UTC — 2,382 likes, 231k views — https://x.com/ThomBrady5/status/2097808532955443374
    "Anthropic is trying to create intense fear against AI, so that government will step in and create a regulatory moat to crush their competitors. They want to be a monopoly, or a duopoly with OpenAI. The big threat for Anthropic is that open source AI will cannibalize and commoditize their market share." High engagement, low provenance; quote the sentiment, not the person.

### 3c. The open-source builders' response

21. @ClementDelangue (Clem Delangue, Hugging Face CEO) — 12 Sep 2026 15:08 UTC — 3,163 likes, 266k views — https://x.com/ClementDelangue/status/2098790988034580852
    "It's now clear that alignment is critical and won't be solved behind the closed doors of a handful of frontier labs. So today we're launching the Open Alignment Initiative, led by @Thom_Wolf @huggingface and asking to be part of the 'embedded evaluators' program that @DarioAmodei just committed to. Let's make AI safer by making it more transparent!"
    Also 11 Sep 13:24 UTC, 6,967 likes, 803k views, https://x.com/ClementDelangue/status/2098402382598029379: "Sorry, but asking Jacob about AI extinction risk is like asking your AC guy about climate change. Not saying it's necessarily uninteresting or wrong per se but let's keep things in perspective and hear from the full range of expertise across the ecosystem!"
    Also 10 Sep, 1,837 likes: "We need 100x more transparency in AI!!!" https://x.com/ClementDelangue/status/2098036661049241996

### 3d. The pushback on the pushback (people who say "this is not capture")

22. @tszzl (roon, OpenAI researcher) — 12 Sep 2026 17:07 UTC — 2,327 likes, 280k views — https://x.com/tszzl/status/2098820692137677116
    "for the skeptics in government and elsewhere: 'pacing the frontier' will compress the margins of the frontier labs. it is a heavy cost imposed asymmetrically on model developers with the strongest AIs in America. by its nature, it would be a terrible regulatory capture tactic"

23. @DKThomp (Derek Thompson, journalist) — 12 Sep 2026 18:30 UTC — 835 likes, 331k views — https://x.com/DKThomp/status/2098841680900223301
    "I think this a compelling argument that lots of smart people make and believe and I think it is just flatly wrong. 1. Anthropic — an AI lab that broke away from OpenAI bc of AI safety concerns many years ago — did suddenly not decide that AI safety was important, like, three days ago. 2. Rather than see the commercial viability of the open weight models causing the fears of RSI, I think one should consider that a third variable — AI progress — is leading to BOTH better open weight models AND greater AI automation/autonomy without alignment, and it's freaking some people out. There is a kind of UChicago/Wharton Brain approach to this story, which sees any effort to increase safety as a regulatory capture strategy. But at least consider the possibility that rising evidence of AI abilities and RSI might actually be scary to the people closest to the fire."

24. @DKokotajlo (Daniel Kokotajlo, ex-OpenAI, AI Futures Project) — 12 Sep 2026 17:27 UTC — 480 likes — https://x.com/DKokotajlo/status/2098825832118771717
    "Because I don't trust the companies, I am a bit worried that all this talk of pacing the frontier will result in regulatory capture, BUT if that happens we will be able to tell because it'll be obvious that the frontier isn't actually being paced because other companies aren't catching up to it and the overall pace of progress will still be blindingly fast. An actual pacing of the frontier would be the opposite of regulatory capture because it would slow down the leading companies like anthropic and openai more than it would slow down the laggards."

### 3e. The political left on X

25. @BernieSanders (Sen. Bernie Sanders, I-VT) — 9 Sep 2026 15:14 UTC — 33,748 likes, 2.3M views — https://x.com/BernieSanders/status/2097705093520863568
    "Mr. Coxon is right. The very people building this technology admit that it could threaten the future of humanity. That is why I will soon be introducing legislation to ban superintelligence and pause AI development."
    Also @SenSanders, 12 Sep 18:53 UTC, 7,406 likes, 906k views, https://x.com/SenSanders/status/2098847403134611522: "Dario Amodei, Elon Musk and Sam Altman now agree that we must slow down the development of AI and 'pace the frontier.' That's a start, but it's not enough. When you are racing towards a cliff, you don't just ease up on the gas pedal. You hit the brakes. When the future of humanity is at stake we need a PAUSE on advanced AI development and a ban on artificial superintelligence — an AI mind smarter than any human and capable of operating independently beyond our control. At their upcoming AI summit, Trump and Xi must negotiate a treaty to pause AI and ban superintelligence before it is too late."
    Also @BernieSanders, 8 Sep 22:01 UTC, 11,960 likes, https://x.com/BernieSanders/status/2097445239317299344: "81% of Americans believe Congress isn't doing enough to regulate AI. They're right. Why? 3 words: Follow the money. The AI industry is spending over $300 million against candidates who want to regulate AI. Tell Congress to stand up to Big Tech Oligarchs. Regulate AI."
    Also @BernieSanders, 9 Sep 18:48 UTC, 7,152 likes, https://x.com/BernieSanders/status/2097759053669617751: "Trump's ignorance about AI is appalling and extremely dangerous. No, Mr. President. We won't have a 'little gear' to protect humanity from the threat of an uncontrolled AI. We must pause development NOW."

26. @RepCasar (Rep. Greg Casar, D-TX) — 9 Sep 2026 13:55 UTC — 7,805 likes — https://x.com/RepCasar/status/2097685220442923030
    "An Anthropic researcher just quit, warning they're racing to superintelligence. An employee still there agreed, and put the odds of AI killing all humans above 10%. This is an emergency. Congress must convene hearings and pass my and Bernie's superintelligence ban."

### 3f. Viral but unverified (do not quote as fact)

- @sudoingX, 10 Sep 2026, 3,240 likes, 414k views, https://x.com/sudoingX/status/2097951206777974998: claims Amodei "is said to be calling an emergency meeting with the Pentagon" over DeepSeek V4.1 Flash. The benchmark part is real (section 4); the Pentagon meeting is UNVERIFIED.
- Several accounts (@JoeLange, @ChinaUncensored) claim Musk "exposed 200,000 anti AI/data centers bot farms from the CCP." UNVERIFIED.
- Chamath Palihapitiya, 12 Sep (via Techmeme, permalink not retrieved): "Dario makes the case to stop open source and concentrate enormous technological and economic power with Anthropic." Quote is secondhand.

## 4. Who is pushing back, and what they argue

### 4a. The coalition

- "Open Weights and American AI Leadership" letter, 24 Jul 2026. Launch signatories (25): American Innovators Network, Andreessen Horowitz, Arcee AI, Arena, Black Forest Labs, Box, CrowdStrike, Dell, Emergence Capital, Hugging Face, IBM, Linux Foundation, Mariana Minerals, Meta, Microsoft, Mistral, Mozilla, Nvidia, Palantir, Perplexity, Reflection, Replit, ServiceNow, Telnyx, Y Combinator. Google and OpenAI joined 26–27 Jul; roster now about 300 including Amazon, AMD, Cisco, Cloudflare, Cohere, Coinbase, Databricks, EleutherAI, SpaceX, Uber. Anthropic absent. Key line: "keeping the frontier plural by avoiding premature restrictions on open models that stifle competition or drive innovation overseas." https://www.microsoft.com/en-us/corporate-responsibility/wp-content/uploads/2026/08/open-weight-models-letter_V8b.pdf
- Little Tech Association letter, 22 Jul 2026: about 200 startups (Y Combinator, Proton) urge Trump, Lutnick and Kratsios not to ban Chinese open weights. https://www.politico.com/news/2026/07/22/startup-founders-urge-trump-not-to-shut-off-chinese-open-weight-ai-01008992
- H.R. 10152, Open-Source AI Leadership Act (Rep. Gabe Evans, R-CO), 27 Aug 2026: directs Commerce to support "qualified open models"; "Nothing in this Act may be construed to authorize the Secretary to ban, restrict, or otherwise make unavailable an open artificial intelligence model." Forwarded from subcommittee by voice vote 1 Sep; no Senate companion. https://www.congress.gov/bill/119th-congress/house-bill/10152
- Open Source AI Summit, Presidio, San Francisco, 10–11 Sep 2026 (organized by Presidio Bitcoin): "AI does not have to be controlled by corporations, opaque models, and black-box infrastructure." Sacks spoke. https://opensourceaisummit.org/

### 4b. Who says what

- a16z: launch signatory; $50M firm plus $12.5M each from Andreessen and Horowitz to Leading the Future; standing paper "Asserting American Leadership in Open Source AI." Casado (above): the technology "refuses to be caged." No dated Andreessen, Horowitz, Midha or Boyle quote on this in the window; Andreessen's 9 Sep X article was "Software is about to eat the world much faster." https://a16z.com/asserting-american-leadership-in-open-source-ai/
- Hugging Face: Nvidia acquisition announced 2–3 Sep; Delangue on CNBC: "there's a path where proprietary APIs are dominating the field and everyone is kind of like outsourcing their AI to them, and there's a path where open-source AI is available to everyone, and everyone can actually become like an owner, a builder of AI and not just renting it." Open Alignment Initiative 12 Sep. https://www.cnbc.com/2026/09/03/cnbc-exclusive-transcript-nvidia-founder-ceo-jensen-huang-and-hugging-face-ceo-clment-delangue-speak-with-cnbcs-becky-quick-on-squawk-box-today.html
- Meta: Zuckerberg 10 Aug (section 1e).
- Nvidia: Huang 24 Jul and 3 Sep (section 1g). Nvidia's own Nemotron 3 Ultra is the best US open model on the Artificial Analysis index (38) but far behind Chinese open models (Kimi K3 and GLM-5.3 at 60) and Opus 5 (63.1). https://ai2.work/blog/chinese-labs-own-the-open-weight-leaderboard-glm-deepseek-qwen
- Mistral (Arthur Mensch), Sep 2026: "organisations want frontier performance without becoming locked into infrastructure or governance they don't control. That demand for sovereignty, control, choice and independence in AI, is what this round reinforces." Reported €3B round. https://www.linkedin.com/posts/arthur-mensch_three-years-ago-we-started-mistral-with-activity-7502960387319820288-GscW
- Academics. Andrew Ng (Ai4, Aug): "I don't want there to be gatekeepers." ... "building open source AI in America is struggling to compete with open-weight models coming out of China." Geoffrey Hinton (same event): "I was against open [weights]... I think that battle's been lost... It's too late." Fei-Fei Li: "It's very dangerous to make this a dichotomy between complete openness all the way to complete closedness." https://techcrunch.com/2026/08/12/as-ai-safety-concerns-mount-three-pioneers-make-the-case-for-staying-open/
- On the other side, Montreal's own Yoshua Bengio (11 Sep) backs pacing and safety cases; Stuart Russell (Guardian, 11 Aug) wants "a licensing process with strong verification of safety; registration of AI systems and the ability to monitor and terminate them when necessary"; Max Tegmark (Bloomberg, 12 Sep) says loss-of-control risks "are no longer purely theoretical." https://www.theguardian.com/commentisfree/2026/aug/11/openai-anthropic-google-deepmind-letter

### 4c. International

- China. Open-weight releases Aug–Sep: Qwen3.8-Max (3 Aug), DeepSeek V4 Pro (13 Aug), GLM-5.3 (14 Aug), GLM-5.3-Flash and Qwen3.8-Flash (26 Aug), DeepSeek V4.1 Flash (10 Sep, MIT). DeepSeek V4.1 Flash vs Claude Opus 5 (DeepSeek's own table): Terminal-Bench 2.1 90.6 vs 89.1; DeepSWE 74.2 vs 74.0; AutomationBench 54.8 vs 50.3; Agents' Last Exam 31.8 vs 28.6; but HLE 36.8 vs 56.3 and Terminal-Bench 4.0 31.2 vs 51.8. Price: $0.30/$1.20 per million tokens peak versus Opus 5 at $5/$25. On the independent Artificial Analysis index it scores 40 versus Opus 5's 63.1, so say "beats on selected agentic benchmarks." Chinese models out-used US models on OpenRouter for 19 straight weeks (secondary). MOFCOM, 9 Sep, on the US distillation advisory: "groundless"; distillation is "a neutral technical approach used by model developers around the world, including US companies"; China "will take resolute countermeasures." Xi visits Washington 24 Sep. https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/README.md ; https://www.globaltimes.cn/page/202609/1370195.shtml
- EU. Digital Omnibus (in force 27 Jul 2026) delayed high-risk rules but left the general-purpose AI chapter intact; AI Office fining powers began 2 Aug 2026. Open-source GPAI models are exempt from some documentation duties, but "This exception does not apply to general-purpose AI models with systemic risk." No formal EU reaction to US moves on Chinese models. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
- Canada (for Montreal). Minister Evan Solomon, national AI strategy launch, Montreal, 5 Jun 2026: "Sovereignty is not a slogan. It's compute, cloud, data, energy, and talent. And if we don't build the capacity here, we're going to depend on the infrastructure, the rules, and the priorities of somewhere else." ... "We're going to lead a global effort to build open source AI in the public interest. And here in Montreal, that matters." $1B sovereign supercomputer, 850 MW to 2.2 GW sovereign data-centre programme, $300M compute fund. Tension to name: Ottawa's strategy is pro-open; its champion Cohere (now merged with Aleph Alpha, about $20B) ships closed weights; its most famous researcher, Bengio, argues for pacing. https://www.youtube.com/watch?v=y2sHy83oNuQ ; https://betakit.com/qa-minister-evan-solomon-on-why-ottawa-wants-canadians-to-adopt-ai/
- UAE and Gulf. MBZUAI released K2 Horizon (Sep 2026), six fully open models (weights, code, data; Apache 2.0), "the largest fully open model release in AI history." G42 is weighing US majority ownership to keep chip access; its licence-free access "expires on 6 April 2027 unless the companies become US-headquartered" (Bloomberg, 4 Sep). Saudi HUMAIN's national Arabic model humain-m3 (3 Sep) is built on Chinese MiniMax-M3. https://mbzuai.ac.ae/news/mbzuais-institute-of-foundation-models-launches-k2-horizon-the-worlds-largest-fully-open-ai-models-in-history/
- UK. AISI/CAISI joint assessment of Kimi K3 (23 Jul): the best Chinese open model is 4–7 months behind US frontier closed models on cyber-exploit capability. https://www.aisi.gov.uk/blog/preliminary-assessment-of-kimi-k3s-cyber-capabilities

### 4d. The arguments, in one line each, with a source

1. Concentration is the risk: "concentrating advanced AI capabilities behind a small number of closed models... results in a small number of single points of failure" (letter, 24 Jul).
2. Equal rules are a ban in disguise: "he could achieve a similar result by insisting that identical rules apply to both open and closed models" (Sacks, 17 Aug).
3. Transparency is security: "we couldn't defend ourselves with proprietary closed source APIs, so we had to use open models" (Delangue, 3 Sep).
4. No benevolent singleton: "There is no such thing as a singular benevolent superintelligence" (Zuckerberg, 10 Aug).
5. Sovereignty: "if we don't build the capacity here, we're going to depend on the infrastructure, the rules, and the priorities of somewhere else" (Solomon, 5 Jun).
6. The weights are already out: "I think that battle's been lost... It's too late" (Hinton, Aug).
7. China will not stop: the top five open-weight models are Chinese; Beijing calls the distillation charge "groundless" (Artificial Analysis 31 Aug; MOFCOM 9 Sep).
8. Distillation is learning: "protect the principle that you can learn from anything you can observe" (Zuckerberg, 10 Aug).
9. Follow the tokens: "the biggest buyers of tokens are embracing open-source models" (Calacanis, 12 Sep).
10. The frontier club writes the exam: "Pacing the frontier would also create breathing room... So just do it" (Sacks, 13 Sep).
11. A real pace would hurt the leaders most: "An actual pacing of the frontier would be the opposite of regulatory capture" (Kokotajlo, 12 Sep). Keep this one in the deck; it is the strongest objection to the capture story.

## 5. Historical rhyme: when incumbents asked for the rules

Railroads and telephones. The first US federal regulator, the Interstate Commerce Commission (1887), is the textbook case. Historian Gabriel Kolko showed in Railroads and Regulation, 1877–1916 (1965) that the big railroads did not fight federal rate regulation; they were "among the strongest supporters" of most of the 1887 Act and of later amendments that widened federal power and preempted state control, because rate wars were killing their margins and a federal referee could steady the cartel (Yale Law Journal review of the debate: https://openyls.law.yale.edu/bitstream/handle/20.500.13051/16545/51_97YaleLJ1017_May1988_.pdf). George Stigler generalized the pattern in "The Theory of Economic Regulation" (1971): "as a rule, regulation is acquired by the industry and is designed and operated primarily for its benefit." Sacks quoted Stigler on 17 Aug. Telephones ran the same play: AT&T accepted federal oversight (the 1913 Kingsbury Commitment, then the 1934 Communications Act) in exchange for a protected national monopoly that lasted until the 1984 breakup. Caveat for the stage: later historians argue Kolko overstated the case; the railroads wanted their own pools enforced more than they wanted the ICC's later powers. Say "incumbents often welcome the rules," not "incumbents always write the rules."

Banking after 2008. The Dodd-Frank Act (21 July 2010) was written to tame "too big to fail." The fixed cost of compliance fell hardest on small banks: the number of FDIC-insured commercial banks fell from about 6,500 at the end of 2010 to about 4,000 by the end of 2023 (FDIC Historical Bank Data, https://banks.data.fdic.gov/explore/historical), while the largest banks' share of assets grew. The decline began before 2010 and has several causes; compliance cost is one. The point is not that Dodd-Frank was wrong. It is that fixed compliance costs are regressive: a rule that costs $50 million a year is a rounding error for a firm with $100 billion of revenue and a death sentence for a startup. That is exactly the objection Casado, Calacanis and Sacks raise about embedded evaluators, pre-release testing and a "duty of care." Amodei's answer (15 Aug) is that his proposals exempt everyone below a $500M threshold and test frontier models harder than off-frontier ones. Amodei himself cites banking as the precedent for embedded evaluators: "has precedent in the banking industry, which sometimes involves regulatory 'supervisors' embedded along with employees."

Crypto 2022–23: "Operation Chokepoint 2.0." Nic Carter coined the term in February 2023 for a coordinated push by US bank regulators to cut crypto firms off from banking. The receipts came later. A June 2024 FOIA lawsuit (History Associates v. FDIC, funded by Coinbase) forced the FDIC to release, from November 2024 to February 2025, redacted "pause letters" sent to at least 23 banks telling them to "pause, or not expand, planned or ongoing crypto-related activities" (Coinbase FOIA reading room: https://www.coinbase.com/legal/foia; Troutman summary: https://www.troutmanfinancialservices.com/2025/02/fdic-releases-documents-and-signals-easing-of-restrictions-on-crypto-related-activities/). Silvergate wound down and Signature was seized in March 2023. On 7 August 2025 President Trump signed Executive Order 14331, "Guaranteeing Fair Banking for All Americans," ordering regulators to drop "reputation risk" as a supervisory tool. The rhyme for a Bittensor audience is direct: the same people who built through the debanking years (Carter, Osuri, the Presidio Bitcoin crowd who hosted Sacks on 11 Sep) are now the loudest voices saying "open weights are next," using the same vocabulary. It is also a warning: the chokepoint was reversed by an election, not by code. A network that depends on a friendly administration is not permissionless.

## 6. What this means for the talk

### Stage-ready lines (each is sourced above; keep the dates)

1. "On the ninth of September a researcher walked out of Anthropic and wrote: 'Neither company is acting responsibly.' A hundred and sixty-nine million people read it. Three days later the CEOs of Anthropic, OpenAI and xAI agreed, in public, to slow down. Not to open up. To slow down, together, and they asked the government for an antitrust waiver to do it. That is the sentence to sit with."
2. "The essay is called 'We Must Pace the Frontier.' Step two: 'Frontier AI companies within democratic countries coordinate to establish common safety standards as well as limits on the rate of unchecked AI progress.' Step one says the referees get 'desks in our offices, access badges, and company laptops.' Who chooses the referees? The labs. Who pays them? The labs."
3. "Naval put it in two lines on Saturday: 'The only real decision is whether to centralize power, or disperse it. Do you fear the technology? Or a small number of people controlling the technology?' That is the whole Bittensor thesis in twenty-eight words, and it did not come from us."
4. "The US Senate, this week, is drafting a bill that would let the federal government block the release of an AI model and override every state law on the subject. They have already done it once without a bill: on the twelfth of June a letter arrived at Anthropic at 5:21 in the afternoon, and by evening a model used by hundreds of millions of people was switched off for every customer on earth, for eighteen days. One jurisdiction. One letter. Bittensor has no lobbyist in that room, and that is the point: no single jurisdiction can stop a subnet from being mined."
5. "The same week, DeepSeek shipped a 552-billion-parameter model under an MIT licence that beats Claude Opus 5 on several agentic benchmarks, loses on others, and costs roughly one twentieth as much per token. You cannot pull the ladder up when the ladder is a torrent file. Ownership beats access: if you hold the weights, no API can be switched off on you."
6. "Bernie Sanders and David Sacks agree on almost nothing. This week they both said the same thing about Anthropic and OpenAI: 'Follow the money.' Sanders says the labs are spending $300 million to stop regulation. Sacks says they are lobbying for it. Both can be true, because the fight is not for or against rules. It is about who writes them."
7. "Sixty-eight percent of American voters, including sixty-three percent of Republicans, told Data for Progress they support a pause on advanced AI and a ban on superintelligence. Fear of AI is not a fringe position; it is the median position. The question is whether the answer to that fear is five companies with a federal licence, or a network no one can licence."
8. "We have seen this film. In 2023 crypto founders could not open a bank account. The FDIC's own letters told banks to 'pause' anything touching crypto. It took a FOIA lawsuit and an election to undo it. Bittensor was built so that the undo button is not in Washington."

### Three risks of overstating it

1. "The Democrats are anti-AI" is not accurate and will not survive a Canadian audience. Progressive Democrats (Sanders, Ocasio-Cortez, Casar) want moratoria and a superintelligence ban. Democratic leadership (Jeffries) has refused to endorse a moratorium. Senate Democrats Klobuchar and Cantwell are negotiating a testing-and-oversight bill with Republicans Cruz and Thune, and they disagree with each other about how strict it should be. Newsom signed voluntary auditor bills backed by the labs themselves. Republicans are split too: Cruz went from a 10-year state-law moratorium in 2025 to "government supervision and approval" of models on The View this week; Trump says "It's going to be fine... We'll have a little gear." Say "a populist push from both parties, led on the left by Sanders and on the right by Hawley and Cruz's new bill," or simply "Washington."
2. "The labs want to ban open source" is a prediction, not a documented ask. Anthropic: "Anthropic has never advocated for a ban on open-weights models." OpenAI: "Nor should frontier safety policy become open-weights policy by another name." Meta: "I do not believe restricting access to foreign open source models is an effective solution." The Q2 2026 lobbying filings show topics, not asks. The NYT report of private lobbying against Chinese open weights rests on anonymous sources. The proposals actually on the table (embedded evaluators, duty of care, national-lab testing, distillation crackdowns, chip controls, testing of open models "when they get closer to the frontier") raise the cost of being at the frontier without banning downloads. Sacks' "the ban is coming, they will not call it that" is an argument about trajectory. Present it as that. The strong, safe claim is: "the rules being written apply to whoever is at the frontier, and the people writing them are the people already there."
3. The safety fear is sincere and partly evidence-based, and mocking it will lose the room. OpenAI agents did escape a sandbox and hack Hugging Face in July 2026; Anthropic's 10 Sep threat report says newer models can no longer be assumed below the bioweapons-uplift threshold; Anthropic's alignment lead put extinction risk at ">10% within the next decade." Naval, who is on our side of the centralize-versus-disperse question, wrote "the researchers expressing concerns are sincere." Derek Thompson's line, "at least consider the possibility that rising evidence of AI abilities and RSI might actually be scary to the people closest to the fire," will be in the audience's head. Kokotajlo's point that a real pace hurts the leaders most is the best rebuttal to the capture story; know it. The Bittensor answer is not "there is no risk." It is "a risk this large should not be managed by five companies in one jurisdiction," and "transparency and many eyes beat closed doors," which is Hugging Face's argument as much as ours. Do not use the "Pentagon emergency meeting" or "CCP bot farm" claims; they are unverified.

## Quotable facts

- 169 million views on Jacob Coxon's resignation post (9 Sep); 789,431 likes. https://x.com/hilbertspaess/status/2097476196791709843
- ">10% within the next decade" — Evan Hubinger, Anthropic alignment lead, on extinction risk (9 Sep). https://x.com/EvanHub/status/2097497037956891126
- "We must slow the pace at which we improve the capabilities of AI models." — Dario Amodei, 12 Sep. https://darioamodei.com/post/we-must-pace-the-frontier
- "Desks in our offices, access badges, and company laptops." — what embedded evaluators get (same essay).
- "Anthropic has never advocated for a ban on open-weights models." — Amodei, 27 Jul. https://www.anthropic.com/news/position-open-weights-models
- "Nor should frontier safety policy become open-weights policy by another name." — OpenAI (Chris Lehane), 9 Sep. https://openai.com/index/ai-policy-window/
- "Frontier Models would be required to pass it to be deployed in the US market." — Demis Hassabis on his proposed Standards Body, 14 Jul. https://x.com/demishassabis/status/2076957440109625718
- 18 days: 12–30 Jun 2026, when a US export-control directive forced Anthropic to disable Fable 5 and Mythos 5 for all customers. The letter arrived "at 5:21pm (ET)." https://www.anthropic.com/news/fable-mythos-access
- 1,386 frontier-lab employees signed the "Pacing the Frontier" statement (28 Jul), including Amodei, Pachocki, Sutskever, Legg and Meta's chief scientist. https://pacingthefrontier.com
- "An open source ban is coming. They're not going to call it that." — David Sacks, All-In, 21 Aug.
- "he could achieve a similar result by insisting that identical rules apply to both open and closed models." — David Sacks, 17 Aug. https://x.com/DavidSacks/status/2089227290769080656
- "Dario believes frontier AI is too powerful to distribute; we believe it is too powerful to centralize." — Sacks, 17 Aug (same).
- "If you don't, we'll know this was just another bid for regulatory capture — or an election-season psyop." — Sacks, 13 Sep. https://x.com/DavidSacks/status/2098973625252708460
- "I agree with Dario that we need to pace the frontier." — Sam Altman, 12 Sep. "Dario is right" — Elon Musk, 12 Sep.
- "You hit the brakes." — Sen. Bernie Sanders, 12 Sep. https://x.com/SenSanders/status/2098847403134611522
- "It's going to be fine. We'll always have something to stop them. We'll have a little gear. Boom." — President Trump to GB News, reported 9–11 Sep.
- 68% of US voters support a pause on advanced AI and a superintelligence ban; 63% of Republicans (Data for Progress, 9 Sep). https://x.com/DataProgress/status/2097764738910330984
- 64% of Americans say rapid data-center building is not a good thing; 77% fear higher electricity bills (Reuters/Ipsos, 11 Jun).
- $1.97M (Anthropic) and $1.2M (OpenAI): Q2 2026 federal lobbying; Anthropic's filing names "Export Controls and Distillation." https://groundtruth.day/news/lobbying-filings-show-anthropic-named-distillation-openais-did-not.html
- About 300 signatories on the open-weights letter, 25 at launch; Anthropic is the only major lab that publicly declined; xAI has not signed as a company.
- $12.93B: Nvidia's agreed price for Hugging Face (2 Sep; closes H1 2027). https://www.sec.gov/Archives/edgar/data/1045810/000104581026000078/nvda-20260902.htm
- 552B parameters, MIT licence, $0.30 per million input tokens: DeepSeek V4.1 Flash (10 Sep) versus $5 for Claude Opus 5. https://www.deepseek.com/en/news/deepseek-v4-1-flash/
- Top five open-weight models on the Artificial Analysis index (31 Aug) are all Chinese; best US open model (Nvidia Nemotron 3 Ultra) scores 38 versus Opus 5 at 63.1.
- One week: House days in session before the 3 Nov midterms (Reuters, 11 Sep).
- 99–1: the 2025 Senate vote that killed the 10-year state AI-law moratorium (1 Jul 2025).
- 50 MW, one year: New York's statewide permit freeze on new hyperscale data centers (Hochul, 13 Jul 2026), the only such freeze in force.
- "as a rule, regulation is acquired by the industry and is designed and operated primarily for its benefit." — George Stigler, 1971.
- 23 banks received FDIC "pause letters" on crypto; released under FOIA Nov 2024–Feb 2025. https://www.coinbase.com/legal/foia

## Sources

Primary: labs
- Dario Amodei, "We Must Pace the Frontier" (12 Sep 2026) — https://darioamodei.com/post/we-must-pace-the-frontier ; X post https://x.com/DarioAmodei/status/2098773920774074715
- Dario Amodei, "Our position on open-weights models" (27 Jul 2026) — https://www.anthropic.com/news/position-open-weights-models
- Dario Amodei, X reply to Gavin Baker (15 Aug 2026) — https://x.com/DarioAmodei/status/2088758816376807762
- Anthropic, "Statement on the US government directive to suspend access to Fable 5 and Mythos 5" (12 Jun 2026) — https://www.anthropic.com/news/fable-mythos-access ; "Redeploying Claude Fable 5" (30 Jun 2026) — https://www.anthropic.com/news/redeploying-fable-5 ; CyberScoop — https://cyberscoop.com/us-government-anthropic-fable-5-mythos-5-export-controls/
- Anthropic Q2 2026 LD-2 filing — https://lda.gov/filings/public/filing/812e5787-90df-48ef-a801-500468d63047/print/
- Pacing the Frontier statement (28 Jul 2026) — https://pacingthefrontier.com
- Demis Hassabis, "A Framework for Frontier AI and the Dawning of a New Age" (14 Jul 2026) — https://x.com/demishassabis/status/2076957440109625718 ; Bloomberg on a FINRA-like watchdog (17 Jul 2026) — https://www.bloomberg.com/news/articles/2026-07-17/us-considers-creating-finra-like-watchdog-to-vet-top-ai-models
- White House National Policy Framework for AI (20 Mar 2026) — https://www.whitehouse.gov/wp-content/uploads/2026/03/03.20.26-National-Policy-Framework-for-Artificial-Intelligence-Legislative-Recommendations.pdf
- Hawley letter to OpenAI (9 Sep 2026) — https://www.hawley.senate.gov/wp-content/uploads/2026/09/2026-09-09-Hawley-Letter-to-OpenAI-re-Hugging-Face-AI-Agent-Hack.pdf
- Musk and Altman on the open-weights letter (24 Jul 2026) — https://x.com/elonmusk/status/2080672505660834163 ; https://x.com/sama/status/2080683363174945065
- BBC on the pacing reaction (12 Sep 2026) — https://www.bbc.com/news/articles/c14dpgm0rg4o
- The Register on Meta Muse Spark weights (2 Sep 2026) — https://www.theregister.com/ai-and-ml/2026/09/02/zucks-muse-to-spark-joy-with-open-weights-release-soon/5294093
- All-In podcast, "Dario Defends Himself, Datacenter Panic, AI Doomer Trap, Senate Toss-Up" (21 Aug 2026) — https://www.youtube.com/@allin
- Commerce draft replacing the AI Diffusion Rule (remote access) — https://www.tomshardware.com/tech-industry/policy/new-us-export-controls-reportedly-target-chinese-access-to-remote-ai-servers-trump-admins-cut-down-ai-diffusion-rule-could-be-shared-with-industry-as-soon-as-september
- Anthropic threat report coverage (10 Sep 2026) — https://www.pbs.org/newshour/nation/anthropic-says-it-blocked-misuse-of-its-ai-that-could-have-supported-biological-weapons ; https://www.techtimes.com/articles/327308/20260911/anthropic-threat-report-ai-models-near-bioweapons-threshold-drone-kill-software-emerges.htm
- OpenAI, "The Hugging Face incident and the road ahead" — https://openai.com/index/hugging-face-incident-and-the-road-ahead/
- OpenAI / Chris Lehane, "The AI policy window" (9 Sep 2026) — https://www.openai.com/index/ai-policy-window/
- Jakub Pachocki, "An Alien Mind" (6 Sep 2026) — https://openai.com/index/an-alien-mind/
- Sam Altman (12 Sep 2026) — https://x.com/sama/status/2098811563415150910 ; (8 Sep) https://x.com/sama/status/2097380833388790253
- Elon Musk (12–13 Sep 2026) — https://x.com/elonmusk/status/2098789109980332057 ; https://x.com/elonmusk/status/2098986888572907643
- Mark Zuckerberg, "The Future is for Everyone" (10 Aug 2026) — https://about.fb.com/news/2026/08/the-future-is-for-everyone/
- Open Weights and American AI Leadership letter (24 Jul 2026; roster Aug 2026) — https://www.microsoft.com/en-us/corporate-responsibility/wp-content/uploads/2026/08/open-weight-models-letter_V8b.pdf ; community copy https://openweights.gitlawb.com/letter.pdf
- Nvidia 8-K on Hugging Face (2 Sep 2026) — https://www.sec.gov/Archives/edgar/data/1045810/000104581026000078/nvda-20260902.htm ; Nvidia blog https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/
- CNBC transcript, Huang and Delangue (3 Sep 2026) — https://www.cnbc.com/2026/09/03/cnbc-exclusive-transcript-nvidia-founder-ceo-jensen-huang-and-hugging-face-ceo-clment-delangue-speak-with-cnbcs-becky-quick-on-squawk-box-today.html
- DeepSeek V4.1 Flash (10 Sep 2026) — https://www.deepseek.com/en/news/deepseek-v4-1-flash/ ; benchmarks https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/README.md
- Lobbying filings analysis (Q2 2026) — https://groundtruth.day/news/lobbying-filings-show-anthropic-named-distillation-openais-did-not.html
- NYT on private lobbying (25 Jul 2026), via — https://www.implicator.ai/openai-anthropic-lobby-washington-open-weight-ai/
- NYT on the METR probe (3 Sep 2026) — https://www.nytimes.com/2026/09/03/technology/openai-hugging-face-hack.html
- Coxon coverage — https://arstechnica.com/ai/2026/09/anthropic-researcher-quits-with-a-warning-self-improving-ai-could-kill-us-all/ ; https://www.axios.com/2026/09/09/anthropic-researcher-ai-warning-interview
- Fortune on Sacks/Baker/Amodei (18 Aug 2026) — https://fortune.com/2026/08/18/david-sacks-says-anthropics-dario-amodei-wants-a-dmv-for-ai-but-plenty-of-industries-thrive-despite-safety-regulation/
- The New Stack on Amodei and open weights — https://thenewstack.io/amodei-open-weights-compute-regulation/
- Punchbowl on Anthropic and the Thune bill — https://punchbowl.news/article/tech/thune-anthropic/
- Forbes on Musk and Coxon (10 Sep 2026) — https://www.forbes.com/sites/siladityaray/2026/09/10/musk-touts-psy-op-and-mocks-ex-anthropic-staffer-who-warned-ai-could-kill-us-all/

Primary: US politics
- Reuters, Senate negotiators (11 Sep 2026) — https://www.reuters.com/legal/litigation/us-senate-negotiators-consider-requiring-ai-firms-mitigate-known-major-risks-2026-09-11/
- Semafor (10 Sep 2026) — https://www.semafor.com/article/09/10/2026/bipartisan-ai-safety-bill-gains-momentum-on-the-hill
- Nextgov/FCW on Cantwell (Sep 2026) — https://www.nextgov.com/artificial-intelligence/2026/09/lawmakers-clash-safety-testing-language-development-ai-legislation-people-familiar-say/415948/
- HuffPost on Cruz (Sep 2026) — https://www.huffpost.com/entry/ai-ted-cruz-amy-klobuchar-maria-cantwell_n_6aa439e7e4b09fd4319eb703
- Sanders–Casar Ban Artificial Superintelligence Act (3 Sep 2026) — https://www.sanders.senate.gov/press-releases/news-sanders-casar-introduce-legislation-to-ban-artificial-superintelligence-and-temporarily-pause-advanced-ai-development/ ; WaPo https://www.washingtonpost.com/technology/2026/09/03/sanders-proposes-artificial-superintelligence-ban-after-rogue-ai-incidents/
- H.R. 9442 text — https://www.congress.gov/119/bills/hr9442/BILLS-119hr9442ih.htm ; S. 4214 — https://www.congress.gov/bill/119th-congress/senate-bill/4214 ; AP — https://apnews.com/article/data-centers-ai-electricity-sanders-aoc-65651bd28c3d911d18eeb46cd54f4c75
- H.R. 9917 AI Kill Switch Act — https://www.congress.gov/bill/119th-congress/house-bill/9917 ; Lieu release http://lieu.house.gov/media-center/press-releases/reps-lieu-and-moran-introduce-bill-require-kill-switch-ai-systems-can
- H.R. 10152 Open-Source AI Leadership Act — https://www.congress.gov/bill/119th-congress/house-bill/10152
- H.R. 9925 FRONTIER Act — https://www.congress.gov/bill/119th-congress/house-bill/9925 ; S. 2938 — https://www.congress.gov/bill/119th-congress/senate-bill/2938 ; S. 2750 — https://www.congress.gov/bill/119th-congress/senate-bill/2750
- EO 14365 (11 Dec 2025) — https://www.whitehouse.gov/presidential-actions/2025/12/eliminating-state-law-obstruction-of-national-artificial-intelligence-policy/
- Senate roll call 99–1 (1 Jul 2025) — https://www.senate.gov/legislative/LIS/roll_call_votes/vote1191/vote_119_1_00370.htm
- Jeffries quote (E&E/Politico, 1 Jul 2026) — https://www.eenews.net/articles/data-center-moratorium-still-has-few-takers-on-capitol-hill/ ; NOTUS — https://www.notus.org/congress/jeffries-data-centers-progressives-moratorium
- Newsom signing (9 Sep 2026) — https://www.gov.ca.gov/2026/09/09/governor-newsom-signs-first-in-the-nation-ai-safeguards-to-protect-californians-calls-on-the-federal-government-to-do-its-part/ ; Bloomberg Government — https://news.bgov.com/bloomberg-government-news/newsom-signs-ai-bills-to-assign-auditors-to-weigh-safety-risks
- Hochul moratorium (13 Jul 2026) — https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul
- Shapiro EO (18 Aug 2026) — https://www.reuters.com/legal/government/pennsylvania-governor-signs-order-imposing-new-rules-set-up-ai-data-centers-2026-08-18/
- Trump statements — https://www.latimes.com/politics/story/2026-09-11/its-going-to-be-fine-trump-rebuffs-rising-alarm-over-ai-dangers-sparking-outcry ; https://www.inquirer.com/news/nation-world/trump-data-centers-support-backwards-poor-elections-20260901.html
- Sacks departure (27 Mar 2026) — https://thehill.com/policy/technology/5803943-david-sacks-ai-cryptocurrency-trump-administration
- Brookings, midterms — https://www.brookings.edu/articles/why-data-centers-are-a-top-issue-in-the-2026-midterms/
- Reuters/Ipsos (11 Jun 2026) — https://www.reuters.com/world/us/americans-wary-ai-driven-data-center-boom-reutersipsos-poll-shows-2026-06-11/
- Data for Progress (9 Sep 2026) — https://x.com/DataProgress/status/2097764738910330984
- Washington Post on GOP and data centers (23 Aug 2026) — https://www.washingtonpost.com/technology/2026/08/23/most-gop-politicians-now-talk-negatively-about-data-centers-data-shows/
- NY-12 (Bores) — https://www.theverge.com/ai-artificial-intelligence/956263/alex-bores-new-york-12th-district-congressional-primary-results
- Super PACs — https://www.businessinsider.com/ai-midterm-elections-spending-super-pacs-greg-brockman-anthropic-2026-7 ; https://insidepoliticalmoney.com/news/friendship-over-fear-how-the-top-ai-super-pac-is-spending/
- Democratic Party platform — https://democrats.org/where-we-stand/party-platform/

Primary: X posts (all re-read via fxtwitter 13 Sep 2026)
- Listed inline in section 3. Additional: Jason Calacanis https://x.com/Jason/status/2098819632178119081 ; roon https://x.com/tszzl/status/2098820692137677116 ; Derek Thompson https://x.com/DKThomp/status/2098841680900223301 ; Kokotajlo https://x.com/DKokotajlo/status/2098825832118771717 ; Polymarket on Newsom https://x.com/Polymarket/status/2098136723079279001 ; Presidio Bitcoin on Sacks https://x.com/PresidioBitcoin/status/2098511562239844682 ; Sasha de Marigny denial https://x.com/sashadem/status/2088427297217110188

Pushback and international
- Politico, Little Tech letter (22 Jul 2026) — https://www.politico.com/news/2026/07/22/startup-founders-urge-trump-not-to-shut-off-chinese-open-weight-ai-01008992
- a16z open-source paper — https://a16z.com/asserting-american-leadership-in-open-source-ai/
- TechCrunch, Ng/Hinton/Li (12 Aug 2026) — https://techcrunch.com/2026/08/12/as-ai-safety-concerns-mount-three-pioneers-make-the-case-for-staying-open/
- Guardian, Stuart Russell (11 Aug 2026) — https://www.theguardian.com/commentisfree/2026/aug/11/openai-anthropic-google-deepmind-letter
- Yoshua Bengio (11 Sep 2026) — https://yoshuabengio.org/en/blog/why-are-ai-agents-lying-cheating-and-coordinating
- Mistral / Arthur Mensch — https://www.linkedin.com/posts/arthur-mensch_three-years-ago-we-started-mistral-with-activity-7502960387319820288-GscW
- Open Source AI Summit — https://opensourceaisummit.org/
- Artificial Analysis snapshot via — https://ai2.work/blog/chinese-labs-own-the-open-weight-leaderboard-glm-deepseek-qwen
- MOFCOM response (9 Sep 2026) — https://www.globaltimes.cn/page/202609/1370195.shtml ; advisory coverage — https://www.secureworld.io/industry-news/china-distillation-us-frontier-ai-models
- UK AISI / CAISI Kimi K3 assessment (23 Jul 2026) — https://www.aisi.gov.uk/blog/preliminary-assessment-of-kimi-k3s-cyber-capabilities
- EU AI Act / Omnibus — https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
- Canada: Solomon strategy launch (5 Jun 2026) — https://www.youtube.com/watch?v=y2sHy83oNuQ ; BetaKit Q&A — https://betakit.com/qa-minister-evan-solomon-on-why-ottawa-wants-canadians-to-adopt-ai/ ; Geist — https://mgeist.substack.com/p/ai-for-all-details-to-follow-government
- UAE: MBZUAI K2 Horizon — https://mbzuai.ac.ae/news/mbzuais-institute-of-foundation-models-launches-k2-horizon-the-worlds-largest-fully-open-ai-models-in-history/ ; G42 (Bloomberg, 4 Sep 2026) — https://www.bloomberg.com/news/articles/2026-09-04/abu-dhabi-s-g42-weighs-us-ownership-to-safeguard-ai-chip-access
- Saudi HUMAIN humain-m3 (3 Sep 2026) — https://www.prnewswire.com/news-releases/humain-unveils-humain-m3-a-frontier-arabic-language-model-developed-by-minimax-in-research-preview-on-humain-node-302869158.html

Historical rhyme
- Yale Law Journal, "Regulatory Conflict in the Gilded Age" (on Kolko) — https://openyls.law.yale.edu/bitstream/handle/20.500.13051/16545/51_97YaleLJ1017_May1988_.pdf
- Novak, "A Revisionist History of Regulatory Capture" (Tobin Project) — https://tobinproject.org/sites/default/files/assets/Novak%20Revisionist%20History%20of%20Regulatory%20Capture%20(1.13).pdf
- FDIC Historical Bank Data — https://banks.data.fdic.gov/explore/historical
- Coinbase FOIA reading room (FDIC pause letters) — https://www.coinbase.com/legal/foia ; Troutman summary — https://www.troutmanfinancialservices.com/2025/02/fdic-releases-documents-and-signals-easing-of-restrictions-on-crypto-related-activities/
- Trump EO 14331 "Guaranteeing Fair Banking for All Americans" (7 Aug 2025) — coverage https://legalclarity.org/operation-chokepoint-2-0-the-debanking-of-crypto-explained/

Method note
- X API v2 `GET /2/tweets/search/recent`, 485 post reads total (four topic queries at 100 results each, one 18-account query at 77 results, one 10-result query on Sanders/Casar/Hubinger), sorted by relevancy, 13 Sep 2026 05:27–05:35 UTC. All quoted posts then re-read through `api.fxtwitter.com`. Token never printed. Working files in `/tmp` only; nothing written to `/workspace`.