Table of Contents
The NSA is reportedly testing the highly restricted Anthropic Mythos Preview AI model, exposing a deep rift within the US government over artificial intelligence procurement. While the Defense Department has actively pushed to block Anthropic since February over supply chain and trust concerns, the National Security Agency is quietly utilizing the company's most potent tool. According to a recent Axios report, the demand for advanced cybersecurity capabilities is currently overriding the Pentagon's internal blacklist.
The Anthropic Mythos Preview is not a standard consumer chatbot. Due to its advanced offensive cyber capabilities, Anthropic has kept the model on a remarkably short leash, limiting access to approximately 40 select organizations. Sources indicate that agencies with access are primarily deploying Mythos to scan their own internal systems for exploitable vulnerabilities, making it a high-stakes defensive asset.
The friction between Anthropic and the Pentagon stems directly from usage policies. Defense officials previously demanded that Anthropic make its Claude models available for all lawful military purposes. However, Anthropic firmly resisted allowing its technology to be used for mass domestic surveillance or autonomous weapons.
This ethical stance led some military officials to label the company as unreliable, a claim Anthropic strongly disputes. Despite this tension, a recent meeting between Anthropic CEO Dario Amodei, White House Chief of Staff Susie Wiles, and Treasury Secretary Scott Bessent focused on the broader security posture of Mythos in government, with both sides describing the talks as productive.
The Future of Government AI Procurement
The NSA's deployment of the Anthropic Mythos Preview highlights a critical reality in modern cyber warfare: operational necessity will almost always trump bureaucratic boycotts. When a tool proves exceptionally capable at identifying zero-day vulnerabilities or securing critical infrastructure, intelligence agencies cannot afford to ignore it, regardless of the Pentagon's procurement politics.
The productive meeting between Dario Amodei and top White House officials suggests a potential thaw in relations. If Mythos transitions from an isolated NSA exception to a broader government standard, it will set a massive precedent. It proves that AI companies can successfully hold the line on ethical boundaries - like refusing autonomous weapon integration - while still securing highly lucrative federal cybersecurity contracts.
