Could This Secret AI Model Change Cybersecurity Forever?

AI startup Anthropic has said its advanced AI model, Claude Mythos, identified more than 10,000 critical cybersecurity flaws across what it described as the ‘most systemically important software in the world’.

The company had announced Claude Mythos in April but chose not to release it publicly due to concerns over potential misuse.

Instead, access was limited to selected partners, including Amazon and Google, through a programme called Project Glasswing.

Reports suggest several governments, including India, have raised concerns over the concentration of such advanced AI capabilities and demanded fair access to the system.

Anthropic claimed that companies participating in Project Glasswing discovered hundreds of critical vulnerabilities using Mythos, with some reporting a more than tenfold increase in bug detection rates.

Cloudflare said it identified 2,000 bugs in its critical systems, including 400 classified as high or critical severity. The company reportedly found Mythos’s false positive rate better than that of human testers.

Mozilla said it found and fixed 271 vulnerabilities in Firefox 150 using Mythos Preview, significantly more than those discovered in Firefox 148 with Claude Opus 4.6.

Anthropic also said the UK AI Security Institute became the first organisation to complete both of its cyber range simulations end-to-end using the model.

Anthropic said major technology firms, including Microsoft, Oracle and Palo Alto Networks, were now patching vulnerabilities at a much faster pace with AI assistance.

The company added that it would work with the US and allied governments to expand controlled access to Claude Mythos, while maintaining that public release remains unsafe due to insufficient safeguards against misuse.

Anthropic is also reportedly preparing for a new fundraising round that could value the company at nearly $900 billion.

Also Read: What If Your Apple Watch Could Warn You About Sleep Disorders? Apple Says It Can