Vulnerability Found In MyChart By Anthropic Model
Epic halted development of new features in their MyChart system (which contains information on over 320 million patients) for approximately six weeks while they fixed security issues discovered by Anthropic’s Mythos model.
The MyChart system is an electronic medical record database containing information on over 320 million patients. Recently, it suffered from some security issues. Epic Systems, the makers of MyChart, stopped developing new features in their system for approximately six weeks while they addressed some vulnerabilities. While that’s interesting enough on its own, what really got me was the way they discovered those vulnerabilities. As reported by The New York Times, Epic’s chief security officer, Sterling Martin, said that the vulnerabilities were identified by Anthropic’s Mythos model. For certain clients, these vulnerabilities could allow unauthorized individuals to view patient information without generating any logs. Epic’s CEO, Judy Faulkner, has stated that they plan to prioritize enhancing the security of their product moving forward.
This is fascinating because instead of being discovered by a typical penetration testing firm, the vulnerabilities were found by an advanced AI model. I use Claude Code every day, and I can tell you that these models are very capable of systematically exploring various scenarios beyond just coding. So if you’re working on a project involving personal data, you might want to run a similar test to ensure nobody beats you to finding your flaw.
Source: techcrunch.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides

Author
Evgenii Arsentev
PhD · Founder & CEO, Dad4Kids
Articles · Latest articles