Anthropic: Mythos Finds More than 10,000 Software Flaws in First Month
Independent checks validated 90% of assessed findings as real, while Anthropic said the tool is now outpacing verification and patching.
- Anthropic's Mythos Preview uncovered over 10,000 high- or critical-severity vulnerabilities in under two months, marking one of the first large-scale accountings of what a frontier AI model can accomplish when applied to widely used code.
- Since Mythos Preview's release, roughly 50 organizations each discovered hundreds of vulnerabilities, with several reporting their bug-finding rate increased by more than a factor of ten; the bottleneck has shifted from discovery to verification, disclosure, and patching.
- Cloudflare identified 2,000 bugs with 400 rated high- or critical-severity, while independent security research firms confirmed over 90% of 1,752 assessed findings as valid and over 62% as high or critical; Mozilla found 271 vulnerabilities in Firefox 150, more than 10 times prior results.
- While Mythos excels at finding vulnerabilities, human capacity to triage, report, and deploy patches remains the limiting factor; open-source maintainers contended with AI-generated bug reports, though Anthropic disclosed 1,129 cases without further vetting, estimating 175 as high or critical.
- Critics argue the breakthrough may stem from massive compute rather than unique reasoning; academic paper Benchmarking Mythos-Linked Bug Rediscovery found public models like GPT-5.5 rediscovered similar vulnerabilities. Anthropic restricted Mythos-class model access due to misuse risks and released Claude Security in public beta, patching more than 2,100 vulnerabilities in three weeks.
14 Articles
14 Articles
Anthropic: Mythos finds more than 10,000 software flaws in first month
Anthropic said its month-old Project Glasswing initiative has uncovered more than 10,000 high- or critical-severity software vulnerabilities across systemically important code, a finding the company says has shifted the central problem in cybersecurity from discovering flaws to verifying and patching them. The findings, drawn from partner reports and independent evaluations, mark one of the first large-scale accountings of what a frontier AI mod…
Weeks after the launch, Claude Mythos, the new AI model, brought a new view of Internet security. Now, Anthropic has revealed that its tool has already identified more than 10,000 vulnerabilities...
'After one month, most partners have each found hundreds of critical- or high-severity vulnerabilities': Anthropic claims Mythos has found over ten thousand major security vulnerabilities across 'the most systemically important software in the world'
Anthropic illustrates the first months of Mythos Preview, saying it discovered thousands of critical and high-severity bugs.
Anthropic Says Mythos Has Uncovered More Than 10K Vulnerabilities
Anthropic says its Mythos Preview model has ferreted out more than 10,000 cybersecurity vulnerabilities. That’s according to a report issued last week by the artificial intelligence startup on Project Glasswing, its effort to use AI to counter AI-powered cyberattacks. “Progress on software security used to be limited by how quickly we could find new vulnerabilities,” […] The post Anthropic Says Mythos Has Uncovered More Than 10K Vulnerabilities …
Anthropic: Claude Mythos identified 10,000+ software flaws
Anthropic and its Project Glasswing partners have identified more than 10,000 high- or critical-severity vulnerabilities in critical software systems, the company announced in an update on the project’s progress. Mythos identifies thousands of high-severity vulnerabilities In April 2026, Anthropic introduced Claude Mythos Preview, a new large language model that can autonomously find zero-day vulnerabilities and create exploits for them. The com…
Coverage Details
Bias Distribution
- 67% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
