Trump met Altman privately at the GOP convention, then attacked Amodei
The week's AI slowdown fight turned political — and the two sides are now trading positions, pricing risk, and pulling back their own enterprise customers.
Sam Altman asked for the meeting; the President spent Monday attacking the other labs. Trump met privately with OpenAI CEO Sam Altman backstage at the Republican midterm convention on Thursday, according to MS NOW, which cites three people familiar with it; Altman requested it, and the discussion covered AI and its growing power. Two days later Altman wrote that slowing the frontier "will be well worth this cost," and OpenAI reportedly pushed its own IPO out past this year on safety grounds — while on Monday Trump posted that the only guardrail AI needs is a "STRONG AND SMART (High IQ!) PRESIDENT," called the extinction scenario a "HOAX," and singled out Anthropic's Dario Amodei as "pretending to be a perfect little angel." Reuters carried the convention report but noted it could not independently verify it. The read: OpenAI is hedging in both directions — telling Washington it will pace voluntarily while keeping the White House warm — and the pacing camp's unity is now a bargaining position, not a bloc. We covered the administration's answer earlier this week — Trump's answer to the pacing ask: he already has criminal power over the labs.
Cybersecurity was the S&P 500's best sector on Monday, and AI fear was the bid. CrowdStrike rose 14%, Palo Alto Networks 13% and Fortinet 9%, per MarketWatch's close — the top performers in the index on a day AI-linked names in Asia and the US sold off after Amodei, Altman and Musk called for slowing development. Seeking Alpha logged CrowdStrike up more than 11% earlier in the session. The market is not saying AI is over; it is paying for insurance on the failure case, which is exactly what a lab chief saying "we might lose control" prices into a portfolio. This is the second cybersecurity rally the AI-safety cycle has produced — AI threats turned cyber earnings into a 20% stock rally.
Nvidia, Palantir and Booz Allen Hamilton reportedly told staff to stop using frontier models. The Information reported that the companies are scaling back or halting use of public OpenAI and Anthropic models over where their data ends up — prompts, responses, logs and usage metadata, any of which can reveal internal tooling and connected apps. OpenAI and Anthropic both say enterprise contracts do not train on customer data by default, but the report notes Anthropic's June policy change permits some retention for abuse monitoring, and up to 30 days in research cases; zero-data-retention terms, private cloud and air-gapped deployments are gaining ground, with roughly three-quarters of the S&P 500 now treating generative AI as a material risk. That is a demand-side safety story the pacing debate has ignored: the labs' loudest critics in enterprise are data officers, not doomers.
What to watch: whether the White House releases the voluntary pre-release model-access framework it finished in August, or leaves it in a drawer.
Should a lab that asks the government to slow the frontier also be allowed to lobby it in private? Tell us in the comments.
