Healthcare Operations

Epic Deploys Unreleased AI Model to Hunt Security Flaws as Clinical Decision Support Battle Heats Up

The EHR giant is using Anthropic's Claude Mythos through Project Glasswing to scan hundreds of millions of lines of code while embedding UpToDate directly into its AI assistant.

2 min read
HA
Health AI Central Editorial Team
Editorial standards & disclosures
August 25, 2026
2 min read
Share:
Epic Deploys Unreleased AI Model to Hunt Security Flaws as Clinical Decision Support Battle Heats Up

Epic Systems unveiled two strategic AI partnerships at its Users Group Meeting that signal the company's dual focus on fortifying its massive codebase against emerging threats while competing aggressively in the clinical decision support market.

The Verona, Wisconsin-based EHR giant is participating in Anthropic's Project Glasswing cybersecurity initiative, deploying the AI company's unreleased Claude Mythos model to scan Epic's codebase—which spans "several 100 million lines"—for security vulnerabilities. Separately, Epic announced a deepened integration with Wolters Kluwer's UpToDate that embeds the clinical reference database directly into Epic's Art AI assistant and Agent Factory platform.

AI-Powered Security Sweep Promises More Patches

Anthropic first announced 50 Project Glasswing partners in April 2025, later expanding the initiative in June to include organizations across healthcare, power, water, communications, and hardware sectors. Epic's participation means health systems should brace for impact.

Epic Chief Security Officer Stirling Martin told attendees that customers should expect a "higher-than-usual number of urgent security fixes" as AI models identify novel attack paths that traditional security reviews might miss. The warning suggests that Claude Mythos is uncovering vulnerabilities at a scale and sophistication that exceeds conventional code auditing methods.

The deployment represents a significant bet on AI's ability to protect critical healthcare infrastructure, with Epic essentially allowing an unreleased AI model to probe the software that manages clinical workflows for a substantial portion of U.S. hospitals.

Clinical Decision Support Gets Direct EHR Integration

Epic's partnership with UpToDate takes a different approach to AI integration, focusing on clinical utility rather than security. The collaboration makes UpToDate's evidence-based clinical reference content available directly within Epic's Art AI copilot for joint customers, eliminating the need for clinicians to switch between systems when seeking clinical guidance.

UpToDate maintains a database reviewed by 7,600 expert clinicians and employs more than 50 in-house physician deputy editors. The platform serves over 3 million clinician users across thousands of hospitals globally. The integration extends beyond Art to Epic's Agent Factory platform, which enables the development of agentic AI applications.

The timing is notable as the clinical decision support market grows increasingly competitive. Wolters Kluwer launched UpToDate Expert AI in October—a generative AI-powered version of its CDS solution—while newer entrants like OpenEvidence, founded in 2022, are challenging established players. Abridge has reported 300 enterprise health system adoptions of its AI-based decision support solution, demonstrating the market's rapid evolution.

Strategic Positioning in a Crowded Market

The dual announcements position Epic at the intersection of two critical AI applications in healthcare: protecting infrastructure and enhancing clinical decision-making. By integrating UpToDate directly into its AI workflows rather than treating it as a separate reference tool, Epic is betting that seamless access to evidence-based content will drive adoption and differentiate its AI offerings from competitors.

For health systems, the developments present both opportunities and operational considerations. More sophisticated security scanning may improve long-term system integrity but will require increased IT resources to implement urgent patches. Meanwhile, embedded clinical decision support could reduce cognitive load for clinicians, though the ultimate impact depends on how effectively AI mediates the delivery of complex medical information at the point of care.

Sources

  1. More tidbits from Epic UGM: Anthropic's Project Glasswing and a deeper partnership with UpToDate — Fierce Healthcare

Companies in this article

Reporting & accountability

How we maintain accuracy

Health AI Central separates reported facts, analysis, and public-source data. Readers can review our standards, request a correction, or consult the relevant research methodology before relying on a claim.

Share: