AI agents just became an insurance problem.
One bill · 1,213 incidents · one airlock
AI agents just became an insurance problem.
Two US senators propose: companies liable when their AI agents hack.
1,213 incidents coded. No stop in roughly 4 of 5.
An AI agent tries to query a system it was never given.
Not authorized. No one allowed this. Stopped.
Sealed as it happens. Anyone can verify it, offline.
Insurers can’t price what no one can prove.
Check. Stop. Prove.
Regulayer™FailStopThe Witness
Sources: O. Perez, 'The Law of Stop', arXiv 2609.22882; Senators Hawley and Murphy, 1 October 2026.