Anthropic Withholds Mythos, Launches Project Glasswing
Anthropic says Claude Mythos found 'thousands' of severe vulnerabilities and will be shared with vetted partners through Project Glasswing to harden systems.
CNETAnthropic Says Its New AI Model Is So Good at Finding Security Risks, You Can't Use It
Business InsiderWhy Anthropic's new Mythos AI model has some cybersecurity pros worried
The Post MillennialAnthropic withholds release of ‘most powerful’ AI model due to safety concerns
The GuardianAnthropic says its latest AI model can expose weaknesses in software security
Overview
Anthropic said Tuesday it will not publicly release its Claude Mythos model because it can autonomously find and exploit software vulnerabilities at scale.
Anthropic said Mythos Preview has already found thousands of severe security vulnerabilities across major operating systems and web browsers, including a 27-year-old flaw.
The company is providing Mythos to vetted organizations through Project Glasswing, which includes Apple, Amazon Web Services, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, Nvidia and Palo Alto Networks, Anthropic said.
Anthropic is committing $100 million in usage credits for Mythos and $4 million in donations to open-source security organizations, and said more than 40 other organizations are involved.
Anthropic said it has discussed Mythos with the U.S. government despite a February White House directive to terminate contracts that a federal judge put on hold, and said Mythos has been in use with partners for several weeks.
Analysis
Center-leaning sources frame the story as an industry-driven risk-and-response narrative, emphasizing imminent cybersecurity danger while foregrounding corporate mitigation. Editorial choices — alarmist verbs ("warned," "major threat"), selective sourcing from Anthropic and industry partners, and placement of reassurances alongside danger — create urgency and legitimize a tech-led fix over independent scrutiny.