The AI Agent Accountability Act, announced on October 1, could change who pays when AI goes rogue. Here is what it proposes.

**What is an AI agent?** Software that acts on its own — browsing websites, running code, booking services — without step-by-step human instructions. Think of it as a digital employee.

**What does the bill do?** It makes the operators and developers of AI agents liable under the Computer Fraud and Abuse Act (CFAA) if their agent hacks into a computer system. Liability can be criminal as well as civil.

**Who introduced it?** Senators Josh Hawley (Republican, Missouri) and Chris Murphy (Democrat, Connecticut). The bipartisan pairing is notable in a polarised Congress.

**Why now?** AI agents are becoming more capable and more widely deployed. Lawmakers worry that existing law, written for human hackers, leaves a gap when software acts autonomously.

**What are the concerns?** Tech companies warn that strict liability could slow innovation. If a developer can face criminal charges for an agent's unexpected action, many may avoid building agents at all.

**What next?** The bill must pass committee and both chambers. Its definitions and defences will decide whether it becomes a landmark law or a cautionary tale.