Anthropic has embedded its own software engineers inside the US National Security Agency (NSA), a report has said, adding that the tech firm is actively helping the intelligence agency deploy its highly advanced “Mythos” AI model for cyber operations. The ‘undercover’ arrangement is taking place despite the fact that Anthropic is locked in an ongoing legal battle with the Department of War, which includes the NSA.Citing people familiar with the matter, the Financial Times, reported that Anthropic has installed about half a dozen ‘forward-deployed engineers’ directly inside the NSA, and these specialists are tasked with customising the AI models for highly specific national security applications and guiding government operators on how to wield the technology.While it remains unclear if the artificial intelligence (AI) startup’s staff are participating in live, active cyberattacks, an insider close to the situation told the publication that the Mythos software is specifically built to infiltrate the heavily fortified computer networks of foreign adversaries like China and Iran.“The best way to build a good defense is to build a good attack,” a person close to Anthropic was quoted as saying. Defending the deployment, the employee reportedly argued that global rivals are already developing similar automated weapons: “If [Mythos] is not used to build attack agents, adversaries will find a way to do it.”
A legal war with the Pentagon
The partnership is highly unusual because the Pentagon recently labeled Anthropic a “supply-chain risk”, a ‘black spot’ for a domestic US business. The relationship between the government and the AI lab soured after Anthropic attempted to legally restrict the US military from using its commercial “Claude” AI models to conduct mass surveillance on American citizens or to power lethal autonomous drones.Anthropic has since sued the Department of War over the security risk label. Yet, even as lawyers trade blows in court, the Trump administration has quietly pushed forward, embedding Anthropic’s technology across various federal agencies, the report suggests.
Why Anthropic’s Mythos triggered global alarm
The secret deployment comes as Anthropic faces mounting pressure from global regulators and private corporations over the raw power of its technology. When the Mythos model launched in April, it was initially restricted to a tiny, elite group of US organisations. Anthropic recently announced that it is expanding access, distributing Mythos to 150 organisations across 15 different countries.
