Anthropic is expanding its cybersecurity focused Project Glasswing by adding 150 new organizations while preparing to bring Claude Mythos class AI capabilities to a broader audience. Quick Summary – TLDR: Anthropic is adding approximately 150 new organizations to Project Glasswing. The expansion brings the initiative to around 200 total partners across more than 15 countries. Existing partners have already identified over 10,000 high and critical severity vulnerabilities using Claude Mythos Preview. Anthropic says Mythos class AI models could become available to all customers in the coming weeks once additional safeguards are ready. What Happened? Anthropic announced a major expansion of Project Glasswing, its cybersecurity initiative designed to help organizations identify and fix software vulnerabilities using advanced AI. The company is adding around 150 new partners, significantly growing the program from its original group of roughly 50 organizations. The move comes as Anthropic works toward making Claude Mythos class models available to a wider customer base, while continuing to emphasize security controls and responsible deployment. We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries. According to Anthropic, the new wave of participants includes organizations from more than 15 countries. The expanded group brings stronger representation from industries such as power, water, healthcare, communications, and hardware, sectors that are often considered critical infrastructure. Many of the newly added organizations are software vendors and maintainers whose code is relied upon by governments, businesses, and millions of users worldwide. Anthropic said these organizations share a common characteristic: a successful cyberattack against their software could have severe consequences at a national or global level. The company noted that for many participants, a major breach could potentially affect more than 100 million people, highlighting the scale of the risks involved. Reports indicate that the expanded partner list may now include organizations such as Samsung and NATO, further underscoring the program’s growing international reach. Anthropic says Project Glasswing partners have been actively using Claude Mythos Preview to scan large codebases for security weaknesses. The results have been significant. According to the company, participating organizations have already identified more than 10,000 high or critical severity vulnerabilities since the program launched earlier this year. Beyond vulnerability discovery, partners are also using the AI model for a range of defensive cybersecurity tasks, including: Writing software patches. Conducting penetration testing. Performing pre-release security reviews. Automating threat detection and response. Modernizing legacy software systems. Anthropic believes these capabilities can help security teams move faster at a time when cyber threats continue to grow in both scale and complexity. A major theme behind the expansion is Anthropic’s belief that highly capable cyber-focused AI models will soon become widely available. The company stated that within the next six to twelve months, many AI developers could possess models with capabilities similar to Claude Mythos. Anthropic also warned that some future releases may not include strong safeguards against misuse. That possibility could dramatically change the cybersecurity landscape, creating new opportunities for both defenders and attackers. As a result, Anthropic says organizations need to adapt quickly and develop new operating practices, standards, and infrastructure designed for an era of powerful AI-assisted cybersecurity. While Claude Mythos Preview remains limited to approved organizations, Anthropic recently revealed plans to bring Mythos class capabilities to all customers. The company said stronger cybersecurity safeguards are still required before a general release can occur. However, Anthropic reported that progress on those protections is moving quickly. Anthropic said: “Models of this capability level require stronger cyber safeguards before they can be generally released. We’re making swift progress on developing these safeguards and expect to be able to bring Mythos class models to all our customers in the coming weeks. At the same time, Anthropic plans to continue expanding Project Glasswing and its Cyber Verification Program to include more critical infrastructure providers, open source software maintainers, and security researchers. I think this expansion shows how seriously Anthropic views the coming wave of AI-driven cybersecurity challenges. Finding more than 10,000 serious vulnerabilities in a short period is a strong indication that advanced AI can become one of the most valuable tools available to defenders. What stands out to me is that Anthropic is not rushing to release Mythos broadly. Instead, it is trying to build safeguards first while giving critical organizations early access. As AI cyber capabilities continue to improve, the race between attackers and defenders is likely to become even more intense, and initiatives like Project Glasswing could play a major role in shaping that future.

Source: SQ Magazine

Read Original Source →