Anthropic’s Mythos model breached nearly all classified US systems during testing, prompting Trump-era access bans and global service suspensions.
The Trump administration imposed restrictions on Anthropic’s Mythos 5 and Fable 5 AI models after they exposed critical vulnerabilities in US classified systems during Project Glasswing. The models identified flaws within hours, leading to a directive barring foreign nationals from accessing them, according to officials cited by reports.
Anthropic disabled the models globally to comply, drawing criticism from over 100 cybersecurity leaders who argue the move weakens US defense capabilities. The NSA and US Cyber Command confirmed the breaches, with Gen. Joshua Rudd acknowledging the models’ effectiveness in penetrating secure infrastructure.
The incident has intensified tensions between the tech sector and Washington, raising concerns over AI governance and national security risks. No immediate market reaction was reported, but the restrictions could impact AI development and cybersecurity policies.