The US National Security Agency (NSA) is reportedly deploying Anthropic's advanced cybersecurity Artificial Intelligence (AI) model, Claude Mythos, despite an ongoing government ban and a 'supply-chain risk' designation from the Department of Defense (DoD). This information comes from a recent Axios report, indicating that the NSA is among roughly 40 organizations currently granted preview access to the powerful AI tool.
The revelation highlights a potential conflict within government priorities, where critical cybersecurity needs may be outweighing existing directives concerning Anthropic's technologies. The Pentagon had previously advised government officials to cease using Anthropic's tools, and the company is currently engaged in legal disputes with the government, including the Trump administration, regarding the ban and broader restrictions on its AI offerings.
Military Broadens Use Amidst Restrictions
Despite the legal and policy hurdles, military entities appear to be expanding their engagement with Anthropic's tools and technologies like Claude Mythos. While the precise applications of the model by the NSA remain unclear, its capabilities are understood to include scanning systems for security weaknesses and identifying potential cyber vulnerabilities.
Access to the Mythos preview is highly restricted due to its potent capabilities and associated risks. Anthropic initially disclosed a list of 12 organizations with access, and the NSA is now believed to be one of the unnamed users. The UK's AI Security Institute also reportedly has access to the model.
Discussions at the White House
In a significant development, Anthropic CEO Dario Amodei reportedly met with White House chief Susie Wiles and Treasury Secretary Scott Bessent on April 17. Discussions during this meeting are said to have focused on the government's use of Mythos, its security protocols, and future deployment plans. This suggests a potential pathway for other government departments, beyond the Pentagon, to integrate Anthropic's AI models into their operations.
Concerns Over AI Misuse
While Claude Mythos promises advanced capabilities for cyber defense, concerns persist regarding its potential for misuse if it were to fall into unauthorized hands. Given these powerful features, Anthropic has limited access to a select group of users as it continues to rigorously evaluate safety measures and meticulously plan any potential broader public release.