America's Most Secretive Spy Agency Eyes AI-Powered Cyberattacks
The U.S. National Security Agency is reportedly moving to integrate Anthropic's latest AI model, Mythos, into its cyber operations — a development that raises serious questions about the use of commercial artificial intelligence in offensive government hacking.
According to a report by TechCrunch, the NSA has been preparing Mythos for deployment in cyberattacks, even as a federal restriction remains in place barring U.S. government agencies from using the San Francisco-based AI company's tools.
What Is Mythos?
Mythos is Anthropic's latest large language model, part of the same Claude family of AI systems the company has been developing since its founding in 2021. Anthropic was founded by former OpenAI researchers, including Dario Amodei, and has positioned itself as a safety-focused AI lab. The company has raised billions in investment and counts Google among its backers.
The apparent contradiction — a safety-first AI company's model being used by an intelligence agency for cyberattacks — is already drawing scrutiny from policy watchers and civil liberties advocates.
A Federal Ban, and a Workaround?
The existence of a federal ban suggests that at some level, policymakers recognized concerns with the NSA using Anthropic's technology — making the reported workaround all the more striking.
Details on how the NSA plans to deploy Mythos remain scarce, but AI models have already shown significant utility in cyber operations: automating vulnerability discovery, generating exploit code, drafting phishing content, and analyzing intercepted communications at scale. Any of these use cases in the hands of a signals intelligence agency would represent a meaningful escalation of AI-enabled offensive capability.
Broader Implications for AI Governance
The report lands at a fraught moment for AI policy. Governments around the world are wrestling with how to regulate — and use — powerful AI systems, while the companies building them navigate competing pressures from national security agencies eager to leverage their tools and a public demanding responsible deployment.
For Anthropic specifically, the report raises uncomfortable questions. The company has repeatedly emphasized safety and responsible AI development as core to its mission. Being associated with NSA cyber operations — particularly if those operations run counter to explicit policy restrictions — could complicate its public-facing commitments.
This isn't the first time major AI companies have faced scrutiny over government and military contracts. Microsoft, Google, and Amazon have all faced employee backlash and public debate over defence and intelligence partnerships in recent years.
What Comes Next
It remains to be seen whether Congress, oversight bodies, or Anthropic itself will respond publicly to the report. The NSA, as a matter of policy, rarely comments on its operational capabilities or procurement decisions.
What is clear is that AI is rapidly becoming a tool of statecraft — and the rules governing how, when, and by whom it can be deployed are struggling to keep pace.
Source: TechCrunch, June 5, 2026