According to a U.S. official and another informed source, the National Security Agency (NSA) has been testing the ability of Anthropic PBC's new artificial intelligence model, Mythos, to identify cybersecurity vulnerabilities in widely used software, including products from Microsoft.

The official and the source indicated that NSA personnel studying the Mythos model were impressed by its speed and effectiveness in detecting potential security flaws. Both individuals requested anonymity as they were not authorized to discuss the matter publicly.

The NSA, a central agency in U.S. cyber espionage operations, has formed an impression that further underscores the significant power of the Mythos model, which has already raised security concerns among leadership in Washington and on Wall Street.

Anthropic initially released Mythos to a limited number of organizations, encouraging them to use the technology to identify software weaknesses in their systems before hackers could exploit them. It was reported on Wednesday that after determining that malicious actors might misuse Mythos, the company decided against releasing the model to the public. The White House also opposed Anthropic's plan to grant access to approximately 70 additional organizations.

Spokespersons for both the NSA and Anthropic declined to comment.