Anthropic is expanding its cybersecurity program, granting a new group of companies access to its Claude Mythos Preview to help them identify software vulnerabilities.
The artificial intelligence lab launched the cybersecurity project, named Project Glasswing, in April of this year. At that time, around 50 partner organizations were given access to the model, including Apple, CrowdStrike, JPMorgan Chase, Microsoft, and Palo Alto Networks.
Anthropic stated on Tuesday that the initial partners, by deploying the model to scan codebases, have so far discovered over 10,000 high or critical severity security vulnerabilities.
The company said this expansion of Project Glasswing represents a new phase, moving it closer to its goal of using AI to enhance the security of all software.
"Following weeks of close collaboration with our Project Glasswing partners, the security industry, open-source software maintainers, and the U.S. government, we are expanding the program to include approximately 150 new organizations," the company stated. It also noted that all organizations must meet specific security requirements before being granted access.
Anthropic indicated that the new organizations are based in over 15 countries, with the majority providing critical infrastructure for many nations. The company did not disclose a specific list of the new entities but said they fill coverage gaps from the first cohort across several industries, such as power, water, healthcare, telecommunications, and hardware.
The company added, "Many of the new partners are suppliers whose codebases are heavily relied upon by numerous organizations and government bodies around the world."
The announcement of Project Glasswing's expansion comes just one day after Anthropic revealed it had confidentially submitted paperwork for an initial public offering, formally setting it on the path to a public listing this fall. The company's latest valuation is approaching $1 trillion.
Comments