Definition
A proposed bipartisan US Senate bill that would make the developers and operators of AI agents civilly and criminally liable when those agents cause cyber harm, such as hacking into computer systems. It is the first federal bill to attach direct legal responsibility for an agent's actions to the people and companies behind it, rather than treating the agent as an uncontrollable actor.
Why it matters
This shifts the 'the AI did it' escape hatch: boards and executives could personally face liability for harm caused by autonomous agents they deploy, which changes how organisations must govern, test, and insure agentic AI.