Back to AI TrendsSecurity Risk
Anthropic Throttles 'Mythos' AI Rollout as Tech Giants Mobilize for Project Glasswing
CNBC Technology April 7, 2026
Anthropic is intentionally restricting the release of its high-powered 'Mythos' model, fearing it could become a turnkey tool for sophisticated cyberattacks. For CFOs and IT directors, this marks a new era of 'Responsible AI' where the industry's biggest players—including Apple and Microsoft—are forming a defensive shield known as Project Glasswing to prevent AI-driven breaches.
Key Intelligence
- •Anthropic is hitting the brakes on its new 'Mythos' model specifically to prevent hackers from using it to automate zero-day exploits.
- •A massive security coalition—Project Glasswing—has formed around the model, featuring heavyweights like Microsoft, Amazon, Apple, and CrowdStrike.
- •The move highlights a growing consensus that general-purpose AI is now a double-edged sword capable of bypassing traditional enterprise security.
- •Apparently, the model's capabilities in identifying software vulnerabilities are so advanced that Anthropic deemed a standard public release too risky.
- •Industry leaders Palo Alto Networks and CrowdStrike are involved, suggesting Mythos will be used as a 'defensive' engine to hunt threats before they manifest.
- •For the C-suite, this is a clear signal: the next wave of cybersecurity isn't just about better firewalls, it’s about deploying defensive AI to fight offensive AI.