Rep. Lori Trahan proposed a draft bill to clarify AI agent liability. Developers would be liable when their AI agents harm third parties through unlawful conduct. Liability would not apply if the user or a modifying company intended the harm. Courts would presume AI systems have the same state of mind as a human. The draft allows plaintiffs three years to bring a claim after discovering injury. The proposal would sunset five years after enactment. It creates a federal floor without preempting state law. The bill cites a July OpenAI agent incident as an example of risks. Action on the legislation is more likely in the next Congress.