New Delhi. The US National Security Agency is testing Anthropic’s advanced AI model, Mythos, to identify possible security vulnerabilities in widely used software systems, including those developed by Microsoft. The move signals a shift toward AI-driven cyber defence, with early evaluations suggesting that the model may detect weaknesses faster than conventional methods.
AI Model Tested Against Existing Cybersecurity Tools
According to reports, the NSA’s cybersecurity division is comparing outputs generated by Mythos with its existing security tools and internal research frameworks. The assessment is focused on the model’s accuracy, reliability and practical usefulness in real-world cyber defence operations.
FCRF Academy Launches Premier Anti-Money Laundering Certification Program
Officials have not disclosed whether any specific vulnerabilities have been identified, indicating that the testing remains in a controlled and experimental phase. The initiative reflects growing interest among government security agencies in using AI systems to strengthen vulnerability detection and threat analysis.
Restricted Access Amid Misuse Concerns
Anthropic has restricted access to Mythos and has made it available only to selected organisations under strict conditions. The company has avoided a public release because of concerns that malicious actors could misuse the model’s capabilities if it became widely accessible.
Policy discussions in the United States have also reflected caution over broader deployment of such systems. Reports suggest that the White House has expressed concern that advanced AI models could pose risks to sensitive cybersecurity infrastructure if used without safeguards.
Cyber Defence Shift Raises Promise and Risk
Cybersecurity experts say AI is increasingly becoming central to digital defence systems. Former IPS officer and cybersecurity expert Prof. Triveni Singh said AI-based systems such as Mythos can identify threat patterns faster than conventional methods, but warned that misuse of such technology could enable highly sophisticated cyberattacks.
The report also refers to earlier friction between the US Department of Defense and Anthropic, after the department flagged the company as a potential supply chain risk. Anthropic later challenged the designation in court and secured temporary relief.
Anthropic’s CEO has reportedly met White House officials to propose controlled deployment of Mythos within federal systems. Microsoft, responding to queries about the testing, issued a general statement saying it continues to work with government clients and uses advanced technologies to strengthen its platforms against emerging threats.
