The National Security Agency (NSA) has been testing Anthropic PBC’s new artificial intelligence model’s ability to find cybersecurity vulnerabilities in popular software, including products from Microsoft, according to a U.S. official and another person familiar with the matter. NSA officials who studied the Mythos model were impressed with the speed and efficiency with which it searched for potential security vulnerabilities, the official and the person said. Both spoke on condition of anonymity because they were not authorized to discuss the matter publicly.
The impression formed by those inside the NSA, the agency at the center of U.S. cyberespionage, is further evidence of the power of the Mythos model that has stoked security concerns among leadership in Washington and on Wall Street.
Anthropic initially released Mythos to only a handful of institutions, encouraging them to use the technology to find software flaws in their systems before hackers did. News reported Wednesday that company executives decided not to release the model to the public after concluding that hackers might exploit it, and the White House objected to the company's plan to provide access to about 70 more institutions.
Spokespeople for the NSA and Anthropic declined to comment.
Related articles:
NSA allegedly has access to Anthropic's high-risk AI model Mythos