Josh Hawley wants companies held liable for recklessly designed AI agents and users for reckless deployment. The US senator explained his proposal in a September 29 op-ed also published by his office.
He also wants criminal accountability when companies know their agents can commit crimes and fail to provide reasonable safeguards. For users, he describes knowingly deploying an agent and allowing it to commit crimes.
Hawley separately calls for mandatory pre-release testing of advanced models for national-security risks. He expects clear liability to create stronger incentives for safeguards.
The article announces legislation; it is not itself a law or evidence of adopted new duties. AI Incidents records this policy position as a risk signal. The expected effect of the rules remains Hawley's assessment.