A new tracking effort reveals Anthropic's Mythos AI access remains tightly controlled among infrastructure, security, finance, and government institutions. The monitoring effort follows named access, reported use, policy negotiations, and public institutional responses to track who controls the most powerful AI system.

Recent developments include a confirmed security incident where unauthorized users accessed Claude Mythos Preview through a third-party vendor environment. An Anthropic spokesperson confirmed the investigation, noting no evidence of core system impact. Meanwhile, Five Eyes nations including the US, UK, and Australia have established access, while the EU and China remain excluded due to export control compliance requirements.

The access landscape shows 12 named partners, overwhelmingly US-based, plus 40+ undisclosed organizations. Critical institutions like the NSA have reported access while the Pentagon remains legally barred despite ongoing negotiations. Financial regulators across Asia-Pacific are actively monitoring Mythos-related cyber risks, with institutions like the Deutsche Bundesbank calling for broader institutional access.