Trump's AI summit may bring talk of balance; New York's hearing puts companies under oath
Zuckerberg, Amodei, Pichai, Brockman and Huang will lunch with Trump on Tuesday, with no bill or licensing regime expected. Five days later, Anthropic, OpenAI, Google and Meta will testify under oath before the New York City Council. Three of the four had to face subpoena threats to attend, and plaintiffs' lawyers and IPO underwriters will read that sworn record first.
Mark Zuckerberg, Dario Amodei, Sundar Pichai, Greg Brockman and Jensen Huang are expected at a White House lunch with President Trump and House Speaker Mike Johnson on Tuesday; Alphabet confirmed Pichai's attendance on Monday, CNBC reported. Johnson set the terms on Monday, telling Fox Business the meeting is about "balance," Reuters reported: "We do not need a moratorium. We do not need to jump in and hyper-regulate this, because we'll lose the race to China." The meeting is unlikely to produce a federal rule. The more immediate costs for these companies may come through sworn testimony, city ordinances with private rights of action, and lawsuits already filed.
On October 5, the New York City Council convenes its full Committee of the Whole, all 51 members, to hear testimony from Anthropic, OpenAI, Google and Meta on AI safety, the first time any of the four will testify under oath before the council. Meta agreed voluntarily after the council's letters went out in mid-September. Google and Anthropic declined by the September 25 deadline, so Council Speaker Julie Menin authorized subpoenas effective Monday morning. OpenAI and Google agreed to appear on Sunday; Anthropic confirmed late Sunday night, just ahead of the subpoena. SpaceXAI never responded at all and was subpoenaed Monday, a fact first reported by CNBC.
The labs are asking for the slowdown Washington refuses. The leading labs, including Anthropic, OpenAI, Google DeepMind, Microsoft and xAI, have themselves called for slowing the pace of AI development; Amodei this month asked the global AI community to slow its release cadence. Yet in Washington, the political leadership is moving the opposite direction, ruling out a moratorium before the meeting that is meant to discuss safety has even happened. Trump had dinner with Amodei on Sunday night, yet in recent weeks he has called AI fears a "hoax" and a "scam" and says guardrails already exist, including at the Department of Justice. The companies asking for caution are the same companies the City Council had to threaten with subpoenas to get on the record in public.
Trump's talk of guardrails at the Justice Department points to a tool aimed at states and cities, not the labs. He signed an executive order in December 2025 directing the Attorney General to build an AI Litigation Task Force. The department launched it in January 2026 to challenge state AI laws that conflict with federal policy. If New York's private right of action becomes law, the task force is the federal instrument most likely to target it.
That gap matters because sworn testimony creates something a White House photo opportunity does not: a transcript that plaintiffs' lawyers can cite and that has weight in court. The council's own slate makes the stakes explicit. Members are moving a first-in-the-nation whistleblower incentive program, a private right of action for New Yorkers harmed by AI agents, and a requirement for independent third-party validation of AI systems. The council's press materials cite "reports that rogue AI agents have already inflicted damage to government databases," language that will sit in the same hearing record as whatever Anthropic and OpenAI executives say under oath.
That record lands on two companies with the most to lose from it. Anthropic is preparing an IPO reported to target a valuation of about $2 trillion, with timing said to be slipping from October toward after the November midterms. Prospectus lawyers read hearing transcripts before underwriters price a deal. SpaceXAI, formed when SpaceX merged with xAI in February, listed in June and is now worth roughly $2 trillion; it closed a $60 billion purchase of Cursor last month. It already carries a Baltimore consumer protection suit over its Grok image tool and a proposed class action brought by teenagers. Neither company volunteered to be examined in public, and both now have less control over what enters the record than they would have had by agreeing in September.
Other developments on Monday point in the same direction. Florida has asked a court to bar OpenAI from developing new models as part of a child harm lawsuit. Anthropic and OpenAI will also skip an Australian Senate AI hearing on October 1. Across two continents, the companies have attended when compelled and avoided hearings otherwise, a pattern plaintiffs' lawyers can use.
Some industry optimists expect federal preemption to make city rules moot, regardless of this week's testimony. The Senate's July 2025 vote suggests otherwise: it voted 99 to 1 to remove a proposed ten-year moratorium on state AI laws from that year's budget bill. Washington could not centralize AI rulemaking with a friendly Congress and a clean legislative vehicle. Tuesday's balance-focused meeting offers no such vehicle. City and state rules remain in force until Congress passes legislation to supersede them.
Tuesday's meeting will probably end with a statement about balance, transparency and voluntary cooperation, without a federal licensing regime or binding rule. On October 5, four companies will answer questions under oath for the first time, creating a public record that outlasts the meeting. If New York's private right of action passes, the desk expects the Attorney General's AI Litigation Task Force to respond first by suing to challenge the city's rule. The industry is then likely to press Congress for broader preemption of city and state AI liability rules.