Anthropic expands Project Glasswing to 150 partners across 15 nations

Anthropic announced on June 2 that Project Glasswing now counts 150 partners in more than 15 countries, all using Claude Mythos Preview to scan critical infrastructure for serious software flaws.

Context

The security‑focused program, originally launched to identify vulnerabilities in high‑impact systems, is being scaled rapidly. According to The Decoder, the newly onboarded partners span sectors such as power, water, healthcare and communications, where a breach could affect up to 100 million people. TechCrunch adds that the expansion targets “critical infrastructure” and that the effort builds on earlier successes where existing participants uncovered over 10,000 serious vulnerabilities.

Impact

By leveraging Claude Mythos Preview, each partner can run automated scans that surface hidden bugs in software that controls essential services. The sheer volume of findings—more than ten thousand severe issues reported so far—demonstrates both the breadth of exposure and the effectiveness of the AI‑driven approach. Anthropic is also commercialising the remediation process through its Claude Security offering, allowing the company to profit from fixing the flaws it helps discover.

What’s Next

The rollout will continue throughout the year as Anthropic onboards additional organizations and refines the Mythos model for deeper analysis. Future updates are expected to broaden the range of systems examined and to integrate tighter feedback loops between vulnerability detection and patch deployment. Stakeholders in the affected sectors are watching closely, as the program could become a de‑facto standard for proactive cyber‑defence in critical infrastructure.