Project Glasswing: Anthropic's Mythos Preview Triggers Emergency Banking Summit with Powell and Bessent
Anthropic's Claude Mythos Preview — capable of finding zero-days in every major OS and browser — triggered an emergency meeting between Federal Reserve Chair Jerome Powell, Treasury Secretary Scott Bessent, and CEOs of Goldman Sachs, Citigroup, Bank of America, Morgan Stanley, and Wells Fargo. Project Glasswing, the defensive initiative built around Mythos, now involves 12 industry partners and $100M in usage credits for securing critical infrastructure.
Original sourceOn April 7, 2026, Anthropic announced Project Glasswing — a defensive cybersecurity initiative built around its unreleased Claude Mythos Preview model — and simultaneously revealed that the model had already identified thousands of zero-day vulnerabilities in every major operating system and web browser during internal testing.
The revelation was significant enough to trigger a rare emergency convening. Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent privately briefed the CEOs of the five largest US banks — including Goldman Sachs CEO David Solomon, Citigroup CEO Jane Fraser, Bank of America CEO Brian Moynihan, Morgan Stanley CEO Ted Pick, and Wells Fargo CEO Charlie Scharf — on the cyber risk implications. According to Fortune and CNBC reporting, the Treasury Department's technology team is now actively seeking access to the Mythos Preview model to begin autonomous vulnerability hunting across federal financial infrastructure.
Project Glasswing itself is a consortium of 12 organizations: Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. Anthropic is committing $100M in Mythos usage credits plus $4M in direct donations to open-source security organizations. Over 40 additional organizations maintaining critical software infrastructure have been granted access.
The dual-use problem at the heart of Glasswing is stark: the same model capabilities that let Mythos find vulnerabilities at scale also make it potentially useful as an offensive tool. Anthropic privately briefed CISA and the Center for AI Standards and Innovation before the public announcement, warning that the model makes large-scale cyberattacks on American systems "significantly more likely this year." The company's response — restrict access to vetted defenders while using the same capability offensively for bug bounties — is a bet that patching faster than attackers can exploit will hold.
Bruce Schneier's analysis on Schneier on Security called the initiative "necessary but insufficient," noting that Anthropic's position of having built the weapon and now offering to fix the defenses creates an inherent conflict of interest that the security community should watch carefully. The 90-day public reporting commitment on findings and fixes will be the first real test of whether Glasswing's defensive posture is genuine or performative.
Panel Takes
The Builder
Developer Perspective
“The 40+ organizations with access to Mythos Preview and $100M in usage credits represents real defensive firepower applied to real infrastructure. If you work in security engineering at a major institution, understanding what this model can do to your systems is now table stakes — get on the Glasswing waitlist.”
The Skeptic
Reality Check
“Anthropic simultaneously claims Mythos can hack every major OS while leading the effort to fix those vulnerabilities — and asking for trust and access to the most sensitive financial infrastructure in the process. The PR architecture of 'we built the weapon, now let us secure your defenses' deserves sustained critical scrutiny, not just acceptance.”
The Futurist
Big Picture
“This is the moment frontier AI ceased being an abstraction in financial and national security circles. An emergency convening of the Fed Chair, Treasury Secretary, and major bank CEOs over an AI model's capabilities is historically unprecedented. The AI security era just went institutional — expect regulatory frameworks, insurance products, and procurement standards to follow rapidly.”