Mythos saga: NSA lost access to Anthropic tool amid dispute

The ongoing Mythos saga gained a concrete new chapter: a New York Times report that the NSA lost access to Anthropic's Mythos tool amid a dispute between the agency and the company. The story reached the Hacker News front page (231 points, 240 comments), fueling debate about the increasingly tense relationship between AI vendors and the US government over how powerful security-capable models are deployed and controlled.

The development extends a storyline that began earlier in June, when the US government ordered a global suspension of Anthropic's Fable 5 and Mythos 5 models after a jailbreak vulnerability bypassed cybersecurity safeguards — an unprecedented regulatory intervention in frontier-model deployment. That episode triggered fierce developer backlash, with one commenter calling it 'a very dark and very sad day' and 38% of polled developers seeing it as overreach that, applied industry-wide, would 'essentially halt all new model deployments.'

The NSA losing access is the new fact: rather than the government restricting Anthropic, here a government agency is the one cut off, inverting the earlier dynamic and raising questions about who controls access to dual-use security tools. It lands the same week OpenAI launched its competing 'Patch the Planet' and GPT-5.5-Cyber effort explicitly positioned 'to take on Anthropic's Mythos,' making the government-vendor friction a competitive opening for rivals.

The broader theme — government-AI vendor tension over security-grade models — is now a defining issue, intersecting with the US pressing Meta on safety reviews and the Anthropic-Alibaba IP fight. Watch whether the dispute is resolved commercially or hardens into a precedent for how national-security agencies and frontier labs negotiate access.