Epic Deployed Claude to Stress-Test 325M Patient Records, Found Critical Flaws
EricTopol · x · 2026-10-02
Per the NYT, Epic Systems — which maintains 325 million electronic records — used Anthropic's Claude Mythos agent to stress-test its systems and found vulnerabilities that could let hackers gain undetectable access to patient data.
- CEO Judy Faulkner revealed the flaw and a six-week remediation plan at an industry conference; most details were previously unreported.
- The urgency prompted Epic to slow development of some product lines while engineers sprinted to patch the holes.
- Faulkner warned patching and attacks may become a never-ending cycle as AI is used on both offense and defense.
More from Companies & People
- Runway AI Summit closes with Valenzuela reflection, Labs unveils Continuum — runwayml · 2026-10-03
- Beverly Hills schools launch AI Task Force with students shaping AI curriculum — lmoroney · 2026-10-03
- Gumloop founders on making companies AI-native: every employee builds agents, IT keeps control — ycombinator · 2026-10-03
- NSF renews EngageAI Institute with $20M for 3D world models in K-12 learning — mohitban47 · 2026-10-03
- Will McGugan on the non-financial rewards of open source, echoing across dev community — A_K_Nain · 2026-10-03
- Anthropic co-founder Olah reportedly pushed to skip Pope's AI encyclical over machine consciousness — kimmonismus · 2026-10-03