Key Highlights
- On June 13, 2026, the Trump administration directed Anthropic to terminate worldwide access to its Mythos and Fable 5 AI systems
- According to Semafor, authorities suspect a China-affiliated entity obtained unauthorized access to the Mythos model
- American security experts worry China might replicate the technology using “distillation” techniques
- Reports emerged that Fable 5, the consumer-oriented variant of Mythos, was compromised — though Anthropic challenged the assessment’s gravity
- Cryptocurrency firms like Coinbase were cut off from advanced defensive AI capabilities provided through Project Glasswing
On June 13, 2026, Anthropic terminated global access to a pair of its artificial intelligence systems following a Commerce Department mandate rooted in national security considerations.
The discontinued models were Mythos and Fable 5. According to Anthropic’s statement, the company lacked the infrastructure to authenticate user nationality on a real-time, per-session basis via API. Rather than implementing selective restrictions, the firm chose complete removal of both systems.
A Semafor investigation revealed that the shutdown decision stemmed partially from intelligence suggesting a China-affiliated organization had gained entry to Mythos. This exclusive reporting came from journalist Reed Albergotti.
American intelligence agencies expressed alarm that Chinese acquisition of the model could enable duplication of its advanced functions through “distillation” — a technique where a secondary AI system learns by studying outputs generated by the sophisticated original.
[[LINK_START_0]]Anthropic[[LINK_END_0]] clarified that White House officials never mentioned Chinese access issues during their direct communications. Instead, those conversations centered on Fable 5 and allegations of its security being compromised.
Capabilities of the Mythos System
Mythos debuted in April 2026 as a cutting-edge AI designed specifically for cybersecurity applications. Anthropic deemed the technology too hazardous for widespread public distribution.
The system operates autonomously to discover and leverage previously unknown software security flaws. Documentation from Anthropic indicates it uncovered thousands of critical zero-day vulnerabilities across leading operating systems and web browsers.
Among its discoveries was a 27-year-old remote crash exploit in OpenBSD, and it demonstrated 83.1% precision on the CyberGym vulnerability assessment. Anthropic’s analysis suggested the model could condense security work that typically requires months of expert analysis into mere minutes.
Distribution was restricted to approximately 12 designated corporate partners and more than 40 institutions participating in Project Glasswing, a defensive security initiative.
This incident marks not the first security lapse. Previously, a Discord community reportedly maintained unauthorized Mythos access for two weeks before Anthropic detected and terminated the connection.
The Fable 5 Security Compromise and Cryptocurrency Ramifications
Fable 5 represents a public-facing iteration of Mythos equipped with enhanced safety protocols. White House AI adviser David Sacks stated on X that Anthropic CEO Dario Amodei received notification about the security breach but minimized its significance.
Anthropic contested this characterization, asserting the compromise was limited in scope and involved vulnerabilities already present in other publicly available models.
Sacks characterized the situation as “serious but easily resolved” and indicated that responsibility rests with Anthropic to address the matter.
The access suspension carries direct consequences for cryptocurrency and blockchain security operations. Coinbase had reportedly begun negotiations to obtain Mythos access via Project Glasswing for protective security measures.
Prior to the prohibition, Zcash’s founder asserted that Mythos had examined the complete Zcash blockchain infrastructure without detecting any vulnerabilities.
Security operations that depended on Mythos for vulnerability identification now face complete service interruption. Anthropic has not provided a timeline for potential restoration of access.
