Anthropic has confirmed the redeployment of its advanced AI cybersecurity model, Claude Mythos 5, to key U.S. organizations tasked with managing and safeguarding critical infrastructure. This decision follows a comprehensive review by governmental authorities that commenced on June 12, 2026.
Initial Impact and Model Capabilities
Claude Mythos 5 initially gained attention in April 2026 when Anthropic introduced it as a groundbreaking tool in cybersecurity, capable of autonomously identifying software vulnerabilities. The model unearthed numerous high-severity flaws across major systems, including a 27-year-old weakness in OpenBSD and a significant flaw in FFmpeg that persisted for 16 years.
The AI model demonstrated a remarkable 72% success rate in generating functional exploits and linking vulnerabilities on its first try. This represented a significant improvement over its predecessor, the Opus model, which managed no successful attempts.
Project Glasswing and Government Review
In response to its capabilities, Anthropic initiated Project Glasswing, a selective access program involving approximately 200 vetted organizations, including U.S. government bodies and tech giants such as Apple, Amazon, and Microsoft. The initiative saw Anthropic allocating $100 million in Claude usage credits to support these efforts.
On June 12, 2026, Anthropic paused access to Claude Mythos 5 and Claude Fable 5, citing ongoing collaboration with the U.S. government to resolve disruptions and resume usage. This suspension temporarily deprived critical infrastructure defenders of a vital cybersecurity tool.
Resumption and Future Prospects
By June 27, 2026, Anthropic received the green light from the U.S. government to redeploy Claude Mythos 5 to a specified group of U.S. organizations handling critical infrastructure. The company is working swiftly to restore access to these entities while exploring broader availability.
This development has significant implications for the cybersecurity sector. It ensures that critical sectors like energy, healthcare, and finance regain access to cutting-edge AI tools capable of identifying zero-day vulnerabilities swiftly.
The phased reintroduction of Claude Mythos 5 represents a cautious yet determined advancement by both Anthropic and government agencies to utilize advanced AI for national cybersecurity while maintaining control over potential dual-use risks.
As the collaboration between Anthropic and the U.S. government progresses, there is a focus on expanding access to Claude Mythos 5 and potentially reintroducing Claude Fable 5, providing a robust defense against cyber threats.
