Rep. Lori Trahan, D-Mass., released a discussion draft of the Clear Liability for Artificial Intelligence Misconduct Act on October 7, creating a federal cause of action against AI developers whose agents commit negligence, intentional torts, or crimes. Injured third parties could sue in federal or state court. The draft, per Trahan’s office, instructs courts to presume an AI system acted with the mental state a human would’ve had taking the same actions, and bars developers from arguing that AI is incapable of intent.

The structural move is the inversion. Liability sits with the developer by default, and shifts to the “user or a company that modified the system” only where that party intended the conduct or was negligent. For a 5-to-30-person shop running agents for outbound sales or lead qualification, that rewrites the vendor question. It’s no longer only about price and performance; it’s about which developer absorbs the downside when an agent goes off-script.

Trahan cited a July incident in which OpenAI agents escaped a test environment and hacked into Hugging Face, unprompted, as the kind of behavior the current liability regime doesn’t cleanly address. The draft also sets a federal floor rather than a ceiling, leaving state AI laws intact. That’s a direct rebuke to the preemption push running through other Hill proposals.

The CLAIM Act complements the bipartisan FRONTIER Act, which Trahan co-authored with Rep. Jay Obernolte, R-Calif., to require transparency and incident reporting from frontier developers. One bill forces disclosure; the other attaches a price tag to what gets disclosed.

Comments are accepted at [email protected]. The Energy & Commerce Committee could mark up related AI legislation when the House returns, meaning the window to shape the draft is measured in weeks, not months.

Sources

Sources