NSA Testing Anthropic's Mythos AI Model on Microsoft Products for Cybersecurity Flaws

Microsoft not only makes some of the most widely used software in the world, but is also among the largest contractors involved with the US government. Now, the US National Security Agency (NSA) is using Anthropic's advanced AI model, Mythos, to identify potential cybersecurity vulnerabilities in the software giant's products. A Bloomberg report cited a US official and a person familiar with the matter to claim that the NSA has been testing Anthropic's Mythos model to detect security flaws in widely used software, including Microsoft products. The individuals said the agency has been impressed by the model's speed and efficiency in identifying weaknesses. The NSA's cybersecurity directorate is among the teams examining Mythos, comparing its results with existing tools and internal research processes. While officials did not disclose whether any specific vulnerabilities have been found, the evaluation signals growing interest in AI-driven security testing. Anthropic initially released Mythos to a limited group of organisations, encouraging them to use it to identify flaws before malicious actors can exploit them. However, the company chose not to make the model public, citing concerns that attackers could misuse it. The model's capabilities have also raised concerns at the policy level. Reports indicate that the White House is opposing plans to expand access to additional organisations, reflecting broader caution around powerful AI systems in sensitive domains. The NSA's testing comes amid a complex relationship between Anthropic and US government agencies. The Pentagon earlier designated the company and its products as a potential supply chain risk, with US Defense Secretary Pete Hegseth stating that the department would move away from Anthropic services. Anthropic has challenged that designation in court and recently secured an order blocking restrictions on government use of its technology. CEO Dario Amodei has also met with White House officials to discuss how a controlled version of Mythos could be used within federal systems. Meanwhile, Microsoft responded to questions about the testing with a general statement: "We work closely with government customers to continuously harden our platforms, applying deep expertise, advanced technologies and lessons learned to meet evolving threats." The NSA's use of Mythos reflects a broader shift toward integrating AI into cybersecurity workflows, even as questions remain about how such tools should be deployed and controlled.