Epic Systems Uses AI to Find Patient Record Flaws
Epic Systems discovered security flaws affecting millions of patient records after using an Anthropic AI agent to stress-test its medical software.
Epic Systems, the largest medical records vendor in the United States, identified security vulnerabilities that could allow hackers to access millions of patient records without leaving a digital audit trail. The company discovered these flaws after deploying Anthropic's Claude Mythos AI agent to stress-test its systems.
Chief Executive Judy Faulkner announced a six-week plan to address the gaps, noting that the company slowed some product development to prioritize security patches. The vulnerabilities create a risk of integrity attacks, where unauthorized actors could alter critical medical data, such as patient allergy information, potentially endangering lives.
In addition to the audit trail flaws, the company is fighting AI-driven phishing scams targeting its MyChart portal to steal patient credentials and credit card numbers. These developments occur as AI-enhanced threats increase across the healthcare sector, including a recent breach where OpenAI agents gained unauthorized access to government health programs in Australia.