Anthropic engineers reportedly assisting NSA with Mythos model

Anthropic, the artificial intelligence company that produces the Mythos model, has reportedly deployed around half-a-dozen engineers to the National Security Agency (NSA). According to a report by the Financial Times, which cited anonymous sources, the engineers are tasked with helping the US intelligence agency use Mythos, which is Anthropic’s frontier cybersecurity AI model. While the engineers are reportedly assisting the agency with certain applications of the model, it remains unclear whether either the engineers or Mythos are being actively utilized in the agency’s cyber operations. The NSA’s mandate involves collecting intelligence from wiretaps, undersea cables, corporate partnerships, and other clandestine means, as well as conducting offensive cyberattacks on foreign adversaries. It is currently unknown if the deployed engineers or the Mythos model itself are involved in these specific hacking operations.

This engineering support is being provided despite an active federal ban on the use of Anthropic’s technology. Axios previously reported in April that the NSA was using Mythos despite the federal ban. The restriction followed a decision by the Department of Defense—the US government department responsible for military operations—to designate Anthropic as a “supply-chain risk,” which is a formal designation regarding a company’s technology. The Department of Defense reportedly designated Anthropic as a supply-chain risk in retaliation for the company not allowing the government to use its models for mass domestic surveillance and autonomous weapons.

The latest report comes as governments are scrambling to gain access to Mythos. Anthropic has claimed that it had to limit access to the model, fearing that its cybersecurity capabilities could be exploited to discover security flaws and carry out hacks. This has created a situation where the company is attempting to restrict access to its technology even as state agencies seek to deploy it.

When reached by TechCrunch, a spokesperson for the NSA declined to confirm or deny the reporting. Anthropic did not respond to a request for comment.

Why it matters

The report highlights the ongoing tension between Anthropic’s safety restrictions and government demand for its cybersecurity AI model, Mythos, despite previous designations of the company as a supply-chain risk.