Definition
A proposed form of criminal liability that would hold a frontier AI developer responsible if an AI system causes serious harm, unless the developer can show it took reasonable steps to prevent it. It shifts the burden onto the developer, similar to corporate 'failure to prevent' laws in other sectors. It is being considered alongside a dedicated AI safety regulator.
Why it matters
It would create personal and corporate legal exposure for AI companies — a major escalation beyond safety regulators and voluntary standards.