Anthropic Expands Project Glasswing to 150+ Organizations as AI‑Driven Vulnerability Discovery Accelerates

Anthropic is widening the reach of Project Glasswing, its cybersecurity initiative built around the Claude Mythos Preview model, by adding roughly 150 new organizations after several weeks of work with its initial group of partners.

The expansion brings the program into more than 15 countries and extends participation across healthcare, energy, communications, technology, and other critical infrastructure sectors. Anthropic said organizations must meet specific security requirements before gaining access and noted that it plans to broaden the program’s geographic footprint over time.

Project Glasswing launched in April with about 50 organizations using Mythos Preview to scan software for vulnerabilities. Since then, the initiative has uncovered more than 10,000 high‑ and critical‑severity flaws, a volume that has sparked industry and government discussions about how to prepare for a future in which advanced AI‑driven cyber capabilities become widely available. Joe Saunders, CEO of RunSafe Security, said the expansion gives essential sectors a chance to prepare for a new phase of cybersecurity, where the speed of vulnerability discovery increases and the focus shifts to reducing real‑world risk.

Anthropic has repeatedly warned that Mythos‑class capabilities will not remain exclusive for long. The company expects other AI developers to release similarly powerful models within six to twelve months, potentially without the safeguards Anthropic is building. In that scenario, the company says cyberattacks could become more frequent and unpredictable, making it critical for defenders to adapt quickly.

While AI is accelerating vulnerability discovery, Anthropic emphasized that security teams still face the heavy lifting of verifying findings, coordinating disclosures, developing patches, and deploying updates. Jim Sherlock, VP of AI and Cybersecurity R&D at ProCircular, said organizations should expect a surge in security advisories from their vendors and warned that patch pipelines not built for this pace will quickly become overwhelmed. He noted that companies risk accumulating massive backlogs if they cannot process and deploy fixes fast enough.

As part of the expansion, Anthropic plans to make some of the vulnerability discovery tools developed for Project Glasswing available to trusted security teams upon request. The company also reiterated its intention to release Mythos‑class models to all customers once it has developed safeguards strong enough to prevent misuse. Anthropic said it is working as quickly as possible but acknowledged that the necessary protections are still under development.