The condition, outlined in a letter sent to congressional leaders this week, injects a new and potentially divisive demand into a debate that has already consumed Washington for more than a year.
The company’s move represents a strategic gamble. For months, Meta has publicly resisted the Kids Online Safety Act, or KOSA, arguing that its broad language could force platforms to censor content related to mental health or sexual orientation. Now, the company is offering its support for the first time, but only if Congress uses the bill to preempt a growing patchwork of state-level AI laws, particularly in California.
California has emerged as the leading laboratory for AI regulation. The state’s legislature has advanced several bills this year targeting algorithmic transparency, deepfake accountability and the use of AI in hiring and healthcare. Meta and other technology giants have warned that complying with a separate set of rules in every state would be costly and legally chaotic, and they have increasingly pressed for a single federal standard.
The letter, addressed to Senate Majority Leader Chuck Schumer and House Speaker Mike Johnson, argues that KOSA provides a natural vehicle for preemption because it already touches on the design of recommendation algorithms. Meta’s proposal would amend the bill to specify that no state may enact or enforce any law that regulates the use of artificial intelligence by online platforms, effectively freezing state efforts in place.
The demand places Democratic leaders in a difficult position. Many of them have championed KOSA as a long-overdue measure to protect minors from harmful content, but they have also supported state-level AI regulation as a necessary check on the technology’s rapid deployment. Several aides on the Senate Commerce Committee said privately that Meta’s condition amounts to a poison pill designed to kill the bill without appearing to oppose it outright.
Republicans have been more receptive to the idea of federal preemption. Senator Ted Cruz of Texas, the ranking Republican on the Commerce Committee, has long argued that a national approach to AI governance is preferable to a state-by-state system. His office declined to comment on Meta’s specific proposal but noted that the senator has previously introduced legislation that would bar states from regulating AI in interstate commerce.
Consumer advocacy groups reacted with alarm. The Center for Digital Democracy called Meta’s offer a “bargain with the devil,” accusing the company of using children’s safety as a bargaining chip to escape accountability for its AI systems. Other groups warned that even if the bill passed with preemption language, the trade-off would leave states powerless to address the most urgent harms of generative AI, from election disinformation to automated discrimination.
The fate of the bill remains uncertain. Congress is scheduled to break for its August recess at the end of next week, and leadership has not yet scheduled a floor vote on KOSA. With Meta’s support now contingent on a provision that divides both parties, the legislation may be forced to wait until after the November election, when the political calculus on both AI and child safety could look very different.
