OpenAI, Anthropic split on AI regulation
OpenAI is pushing states to copy a common set of AI bills toward federal preemption. Anthropic is funding states to keep passing tougher ones.
OpenAI and Anthropic are running opposite plays on AI regulation, and both are backing them with real money. OpenAI’s chief of global affairs, Chris Lehane, told The Hill the company is chasing what he calls “reverse federalism”: get a handful of favorable state bills passed, get other states to copy them, and use that de facto national standard as the case for eventual federal preemption of stricter state rules. Anthropic is doing the reverse, encouraging states to keep raising the bar. It was the first major lab to back California’s SB 53, and has since endorsed New York’s RAISE Act, Illinois’s SB 315 (which Governor JB Pritzker signed July 6), and a Massachusetts safety proposal it calls the toughest in the country.
The fight already has a price tag. In June’s Democratic primary for New York’s 12th congressional district, groups tied to the two labs spent over $27 million combined on one race: about $19 million from Anthropic-funded Public First Action backing state legislator Alex Bores, and about $8 million from OpenAI-aligned Leading the Future opposing him. Neither side’s pick won; Micah Lasher took the seat and told both companies in his victory speech he would not take cues from “either of you” on regulating AI. Massachusetts is next up for heavy spending from both camps.
Update, July 21: Anthropic gave Public First Action another $20 million, bringing its total support to $40 million. Anthropic framed both rounds strictly as issue advocacy, not campaign spending:
Both of our donations were made exclusively to support Public First Action’s public education and policy mission, and cannot be used to influence the election of any candidate for federal, state, or local office.
The company is funding advocacy for its Advanced AI Framework (government verification of safety claims, mandatory testing of high-risk models, export controls on advanced chips), and tied the new donation’s timing to recent capability jumps, including Claude Mythos Preview’s discovery of thousands of high-severity software vulnerabilities, as evidence for why it wants tougher state rules in place now.
For anyone building on these platforms, “wait for federal law” is not a safe planning assumption right now. Whichever theory wins, state legislatures are writing the actual rules this year, and the two labs’ own money is now a signal worth tracking alongside the bill text. For more on how fast state rules can bite, see state AGs open investigation into OpenAI.