NSA uses Anthropic's AI despite ban

Microchip shortage forced NSA to continue using Anthropic's AI technology despite Pentagon ban. White House approves $9 billion for secret data centers.

See also: Anthropic's Mythos Preview creates functional exploits

The US government is facing a problem it can’t solve publicly. The Pentagon has officially blacklisted Anthropic as a threat to the national security supply chain. But the NSA is using Anthropic’s AI technology because there is no alternative.

White House Chief of Staff Susie Wileshas authorized the National Security Agency to continue using an advanced model from Anthropic, the New York Times reports. The decision was necessitated by a critical shortage of advanced microchips needed to run the cutting-edge AI models on classified networks.

The compromise came alongside a secret $9 billion emergency funding request approved by the White House. The money is intended to help America’s top spy agencies, including the CIA and NSA, secure the high-quality semiconductors needed to run genetic AI in top-secret infrastructure.

The computing power requirements of modern AI have exceeded the expectations of defense experts and congressional committees. Cutting-edge models consume processing power far beyond what secret networks were designed to provide. Since the government cannot secure enough physical microchips, intelligence agencies are unable to deploy or fully test the latest AI tools.

The $9 billion will fund specialized federal data centers built to house Grace Blackwell . These systems require specialized construction with massive electrical power and specialized liquid cooling. They can't operate on typical government computer networks.

See also: Anthropic acquired the startup Stainless

Congress will formally vote on the package. Meanwhile, the White House is redirecting $800 million from other government budgets to immediately begin purchasing computing capacity. The urgency is driven by fears that China will gain an advantage in global intelligence operations.

Military and intelligence agencies rely on AI to analyze millions of intercepted communications, satellite imagery, and data. AI detects anomalies and reveals threats that human analysts would miss. The shortage of microchips that is holding up these tools is being considered a national emergency by the government.

The contradiction is stark. The same government that labeled Anthropic a threat to the supply chain now depends on Anthropic’s models because it lacks the hardware to run alternatives on its own infrastructure. The ban was prompted by concerns about Anthropic’s corporate structure and its relationships with foreign investment. The functional dependency is due to the fact that Claude is one of the most capable logic models available.

Project Glasswing , which gives 50 select partners access to Claude Mythos for vulnerability discovery, has identified more than 10,000 critical vulnerabilities in a month. The company is both the most capable AI security tool available to Western defense contractors and a company the Pentagon considers a supply chain risk.

The microchip shortage extends beyond the intelligence community. The same memory reallocation that is affecting the smartphone market, as Samsung, SK Hynix and Micron redirect drives from consumer electronics to AI, is also limiting the government’s ability to build secret AI infrastructure. The $9 billion request reflects the same problem that is plaguing consumer electronics worldwide.

See also: Anthropic Mythos: Updates the Financial Stability Board on its findings

The situation highlights a structural tension in U.S. AI policy. The government wants to control which AI companies have access to sensitive functions. But cutting-edge AI capability is concentrated in a few private companies. When the hardware to make alternatives work doesn’t exist, the government’s influence over its supply chain disappears. The $9 billion is designed to fill the hardware gap.