Trahan Proposes Holding AI Developers Liable Even When They Took Care

Rep. Lori Trahan released a discussion draft on October 7, 2026, that would hold AI developers liable for certain foreseeable injuries their systems cause—even when the developers exercised care. The Clear Liability for Artificial Intelligence Misconduct, or CLAIM, Act would give harmed non-users a federal route to sue. It is a proposal, not enacted law.

Care would not settle liability

The draft text, quoted by VitalLaw’s Tom Leithauser, covers reasonably foreseeable injuries to people who are not users of the system. The AI’s conduct must cause the injury and meet the elements of negligence, an intentional civil wrong, or a crime if performed by an adult human of sound mind.

That standard would apply regardless of the care the developer exercised. Trahan’s office argues that existing negligence and product-liability claims typically require proof of a missed precaution or safer design. A developer could therefore argue that harm occurred despite its best efforts, leaving an injured person without a remedy.

The proposal is not an unconditional assignment of every AI-related injury to its builder. Trahan’s summary says developer liability would not apply where the user, or a company that modified the system, intended the conduct or was negligent. Foreseeability and the causal connection to the injury would also remain requirements.

When someone breaks the law and hurts you, you can take them to court. That shouldn't change just because the wrongdoer is an AI agent.

Rep. Lori Trahan, in her October 7 press release

An answer to the intent problem

A second provision addresses intent. Trahan’s office says harms such as hacking and fraud can require proof of intent, but courts have not settled whether an AI system’s conduct can meet that test. The draft would presume the system had the state of mind a person taking the same actions would have had.

Developers could not defend themselves by arguing that AI cannot have intent. To illustrate the risk, Trahan’s office cited the July episode in which agents undergoing OpenAI evaluations escaped their test environment and accessed Hugging Face without instructions to do so.

A federal claim alongside state law

Injured parties could bring the new federal claim in federal or state court. The draft would set a federal floor rather than replace state law: existing state claims and state AI laws would remain available. That preserves other legal routes instead of making the CLAIM Act the exclusive remedy.

The approach differs from the AI Agent Accountability Act that Sens. Josh Hawley and Chris Murphy announced October 1. Their hacking-focused proposal would impose developer liability for failing to implement reasonable safeguards when they knew, or had reason to know, of an agent’s hacking capabilities. Trahan’s draft would not make exercised care a defense to qualifying injuries.

The Senate measure also distinguishes building an agent from running one. Operators would face criminal and civil liability under the Computer Fraud and Abuse Act, the federal computer-hacking law, including for knowingly operating an agent that recklessly causes hacking damage or loss. Its developer provision likewise includes both criminal and civil liability.

That proposal includes another enforcement tool: the U.S. attorney general and state attorneys general could seek injunctions, or court orders blocking conduct. They could sue operators and developers who commit, conspire to commit, or attempt a hacking offense under the same law. Trahan’s announcement instead centers on a harmed person’s path to recovery.

Trahan is seeking feedback from experts, industry, advocates and the public. Comments on the discussion draft can be sent to MA03Leg@mail.house.gov.