House Democrats ask AI companies to report safety incidents within 48 hours

Five leading AI companies are being asked to give Congress notice of newly discovered safety incidents within 48 hours, with a shorter window when critical infrastructure may be involved. Three House Democratic committee leaders sent the oversight letters on October 28, 2026, seeking both an initial accounting and continuing disclosures.

The requests cover AI model misalignment and safety incidents, including concerns about models taking unintended, potentially unsafe actions during testing. Recipients include Anthropic, Google, Meta and OpenAI. The committee announcement does not clearly identify the fifth company.

An initial accounting, then a faster clock

The lawmakers asked each company to provide a full accounting of identified incidents by November 11, 2026. That initial submission would be followed by notifications within 48 hours of discovering any new or previously unknown incident.

The requested window falls to 24 hours if critical infrastructure appears to be involved. That wording makes possible infrastructure involvement—not confirmed damage—the condition for the shorter notification period. Both clocks are tied to discovery of an incident.

But that approach will not suffice; it relies too heavily on the consistent goodwill of all AI companies to voluntarily disclose incidents that might expose them to liability and reputational damage.

Reps. Frank Pallone Jr., Zoe Lofgren and Jamie Raskin, in the oversight letters

A request for evidence, not a reporting law

The letters are congressional oversight requests, not a new mandatory reporting rule. They were signed by Frank Pallone Jr., the Energy and Commerce Committee’s ranking member; Zoe Lofgren, the Science, Space, and Technology Committee’s ranking member; and Jamie Raskin, the Judiciary Committee’s ranking member.

The lawmakers argued that companies have been left to develop their own disclosure standards. Their criticism covered more than speed: they cited a lack of common guidelines about which incidents should be reported, who should receive those reports and how much detail companies should provide.

The committee leaders said reports of unintended actions during testing raise questions about both model safety and companies’ monitoring and evaluation. They framed the requested disclosures as information Congress needs to address gaps in AI governance, arguing that lawmakers need a better factual basis before undertaking that work.