Microsoft announced that it will officially join the “Project Glasswing” launched by artificial intelligence company Anthropic and become one of the core partners of the program. Under the agreement, Microsoft will gain access to Anthropic's latest AI model, Mythos, and integrate it into its own cybersecurity defense system to identify and repair potential vulnerabilities in software.
Mythos is currently the most powerful general-purpose large model developed by Anthropic. Although it is not specifically trained for network security, with its excellent programming and reasoning capabilities, it has independently discovered thousands of major vulnerabilities in operating systems, browsers and other software, including security flaws that have been hidden in OpenBSD for 27 years and code problems that have not been detected in FFmpeg for 16 years.
In order to prevent technology abuse, Anthropic has adopted a cautious phased release strategy and will not open Mythos to the public for the time being. It will only provide limited access to partners of the "Glass Wing Project" and about 40 critical infrastructure maintenance organizations. In addition to Microsoft, technology and network security companies such as Amazon, Apple, Google, Nvidia, Cisco, CrowdStrike, and Palo Alto Networks have also joined the program.
Anthropic promised to provide up to US$100 million in usage for the project and donated US$4 million to open source security organizations, requiring participants to share research results. Microsoft said it will use Mythos to conduct in-depth security audits on the Azure cloud platform and Windows systems to improve its response to zero-day vulnerabilities.
Currently, the U.S. Treasury Department, the Federal Reserve and other institutions have held emergency meetings on the new cybersecurity risks that Mythos may bring to assess its impact on the financial system and critical infrastructure.
