The Omniscient Bulletin · 2026-09-16
The Omniscient Bulletin — September 16, 2026
The answers to Dario Amodei's pacing proposal kept coming on Tuesday, and the two that mattered most came from Meta and OpenAI. Mark Zuckerberg said every lab already has the incentive to train safely, and that Meta delayed Muse for months without asking anyone else to. OpenAI confirmed it has been talking to Anthropic and Google about safety dating back to a July proposal by Demis Hassabis, and its policy chief said the antitrust waiver Amodei wanted is unnecessary. Google shipped two new models and rested their frontier safety conclusion on an evaluation of an earlier model. Away from the argument, Amazon said it cannot restore access to data hosted exclusively in Bahrain, or recover data in one UAE zone, New York put a number on what towns should ask per megawatt, and Semafor reported that Washington leaned on a prediction market, which Commerce says never happened.
Tuesday was the day the slowdown argument met its counter-offers. Mark Zuckerberg wrote the longest of them, rejecting coordination on the grounds that alignment is itself a competitive advantage, and disclosing that Meta had delayed shipping Muse for several months on its own. OpenAI went further and said the coordination is already happening, and that it started before the essay: a spokesperson confirmed talks with Anthropic and Google running since Demis Hassabis proposed a US led standards body in July. The company's policy chief said the antitrust waiver Amodei had asked Washington to consider is not needed. Google, meanwhile, shipped its most advanced live dialogue models and told readers of the model card that their frontier safety conclusion rests on an evaluation of a different model. The rest of the day was harder and less rhetorical. Amazon told customers it cannot restore access to data hosted exclusively in Bahrain, and cannot recover data in one United Arab Emirates zone. New York urged towns to ask data center developers for a set sum per megawatt. And Semafor reported that the Commerce Department had quietly ordered a prediction market to unpublish the curve it ran for the price of AI compute, which Commerce says it never did.
Frontier
Zuckerberg answered the slowdown call by saying every lab already has the incentive to train safely, and that Meta delayed Muse for months without asking anyone else to
Mark Zuckerberg answered the pacing proposal on Tuesday night. "Every lab has the responsibility and incentive to move at the pace required to train its models safely," he wrote, and a lab that does not focus on alignment "will fall behind." Meta "delayed shipping Muse for several months to focus on safety and security," he said, and "didn't call for everyone else to do this before we would." He called independent evaluators best practice, and said Meta has committed "the significant majority of compute" to serving people rather than racing toward recursive self-improvement.
Policy
OpenAI confirmed it has been in talks with Anthropic and Google about working together on safety, dating back to a July proposal by Demis Hassabis
An OpenAI spokesperson confirmed to CNBC on Tuesday that the company "has been engaging with other artificial intelligence companies, including Anthropic and Google, about how they can work together to address safety concerns." The discussions have been ongoing since Demis Hassabis's July proposal for a US led standards body modeled on the Financial Industry Regulatory Authority. Amodei had asked Washington to consider waiving antitrust rules so rivals could coordinate; Lehane told reporters at a Washington briefing the same day, Bloomberg reported, that he sees no need for one.
Frontier
Google shipped two new live dialogue models and rested their frontier safety conclusion on an evaluation of an earlier model
Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on Tuesday, and published a model card the same day. It reports no fresh frontier safety evaluation of either. "We evaluated Gemini 3.7 Flash as outlined in our latest Frontier Safety Framework, and we found that it did not reach any Tracked or Critical Capability Levels," it reads. Because the new models show no "meaningful new capabilities or material increases in performance compared to Gemini 3.7 Flash," Google says they are "not likely to reach any T/CCLs."