The HackerOne program invites researchers to find and report cybersecurity jailbreaks in Fable 5 — a straightforward if overdue move for a frontier lab. The more interesting piece is the accompanying Jailbreak Severity Framework, a two-axis scoring system (attack accessibility × potential harm) that Anthropic built alongside AWS, Microsoft, and Google. The explicit goal is giving governments and labs shared vocabulary so that a low-severity jailbreak doesn't trigger a disproportionate regulatory response. That framing is doing real work. Labs have long complained that policymakers conflate party tricks with genuine safety failures; a shared severity rubric, if it gains traction, shifts those conversations onto more precise ground. Whether governments adopt it or treat it as industry self-interest dressed up as governance is the open question.