Project Glasswing: Claude Mythos Has Found 10,000+ Critical Bugs — and Humans Can't Patch Fast Enough

Anthropic announces Claude Mythos has identified 10,000+ high- and critical-severity vulnerabilities as Project Glasswing outpaces human patching capacity

Anthropic disclosed that its Claude Mythos model, deployed through Project Glasswing, has now identified more than 10,000 high- or critical-severity vulnerabilities in critical software systems — a scale of AI-assisted bug discovery with no precedent.

The company and six independent research firms have assessed 1,752 of the high- and critical-severity findings, validating more than 90 percent as true positives. Mythos has scanned over 1,000 open-source projects, flagging 23,019 total issues — 6,202 of them high- or critical-severity.

The bottleneck is no longer discovery speed. Open source maintainers, overwhelmed by the volume of rigorously documented vulnerability reports, have become the limiting factor. Anthropic acknowledged that even with detailed reports delivered to project maintainers, the patching pipeline can't keep pace.

"At present, no company — including Anthropic — has developed safeguards strong enough to prevent such models from being misused and potentially causing severe harm," the company concluded, noting that Mythos-class models will only be made generally available after stronger safeguards are in place.

Sources: Help Net Security, VulnCheck