Anthropic announced Tuesday that its Project Glasswing program will now serve about 150 new organizations in over 15 countries, granting each access to the company’s Claude Mythos AI model. The expansion dramatically widens the pool of critical‑infrastructure operators that can scan their codebases for hidden vulnerabilities.
The AI firm, which filed a confidential initial public offering after a $65 billion financing round that valued it near $1 trillion, says Project Glasswing is designed to locate and remediate severe software bugs before they can be exploited. Claude Mythos, described by Anthropic as its most powerful model yet, can identify thousands of zero‑day flaws over a period of weeks.
In early April, Anthropic rolled out a preview of Mythos to 50 initial partners, including the U.S. government, allowing them to test the model on their own software stacks. Those early adopters helped the company fine‑tune the system before the broader rollout.
The newly added partners span sectors that were under‑represented in the first cohort: power utilities, water treatment firms, health‑care providers, communications carriers, and hardware manufacturers. Anthropic emphasizes that a breach in any of these domains could have catastrophic consequences. The firm estimates that a successful attack on many of the participating codebases could affect more than 100 million people, with serious ramifications for both national and global security.
Countries joining the expanded program include allies of the United States such as Australia, Canada, France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand and South Korea. The list reflects Anthropic’s focus on partners in jurisdictions it deems friendly to U.S. interests.
Among the organizations now receiving Mythos access are identity‑management firm Okta, South Korean tech giants Samsung, SK Hynix and SK Telecom, the NATO alliance headquartered in Brussels, and the European Union’s cybersecurity agency ENISA. These entities maintain codebases that other governments and companies rely upon, heightening the stakes of any potential exploit.
Anthropic’s push comes as the competitive landscape heats up. OpenAI recently introduced GPT‑5.5‑Cyber, a model built specifically for cybersecurity tasks, and has begun testing it with a large group of partners. Anthropic acknowledges that rivals will soon field models with comparable capabilities, prompting the company to race ahead on safety measures embedded within Project Glasswing.
In a blog post announcing the expansion, Anthropic wrote, “What each partner has in common is that a successful attack on their codebase could be catastrophic.” The firm added that it is establishing safeguards to prevent misuse of Mythos while still enabling rapid vulnerability detection.
The broadened rollout underscores how AI is becoming a central tool in protecting the digital foundations of modern societies. By equipping a diverse set of critical‑infrastructure operators with advanced vulnerability‑hunting technology, Anthropic hopes to reduce the window of opportunity for threat actors and reinforce both national and international security frameworks.
As AI‑driven cybersecurity solutions proliferate, the race to balance powerful detection capabilities with robust safeguards will likely shape the next wave of digital defense strategies.
This article was written with the assistance of AI.
News Factory SEO helps you automate news content for your site.