Gottheimer readies AI bill to vet powerful AI models for risk

Rep. Josh Gottheimer is preparing to introduce a bill that would require some AI companies to submit their most powerful new models to the government for risk screening. The legislation targets national security, cybersecurity, and bioterror threats. The bill has not yet been formally introduced, and details including which models would qualify as 'powerful' are left to future rulemaking.

The proposed screening process would involve companies providing access to their models for government evaluation before public release. Gottheimer's office stated that the aim is to catch potential risks early, particularly those that could be exploited by adversaries. The bill does not specify penalties for non-compliance, and it remains unclear which agency would oversee the assessments.

The definition of 'powerful' models is a critical detail that will determine the bill's scope. The timeline for a hearing or vote remains uncertain, but the proposal signals continued legislative focus on AI safety.