Anthropic’s Mythos AI Allegedly Compromised by China-Linked Hackers: What You Need to Know

Key Takeaways

• On June 13, 2026, the Trump administration mandated that Anthropic terminate worldwide access to its Mythos and Fable 5 artificial intelligence systems
• Intelligence sources suggest a Chinese-affiliated organization may have obtained unauthorized entry to Mythos, according to Semafor’s investigation
• American security experts worry China might replicate the technology using “distillation” techniques
• Reports emerged of Fable 5 being compromised through jailbreaking methods — Anthropic challenged these claims’ significance
• Cryptocurrency exchanges like Coinbase experienced disruption to their defensive cybersecurity capabilities via Project Glasswing

On June 13, 2026, Anthropic terminated global availability of two proprietary AI systems following a Commerce Department mandate related to national security threats.

The systems withdrawn were Mythos and Fable 5. Anthropic explained that real-time nationality verification for individual API requests was technically unfeasible, prompting complete removal instead of selective restrictions.

According to Semafor’s exclusive reporting by Reed Albergotti, intelligence suggesting Chinese-connected entities had penetrated Mythos influenced the shutdown decision.

American intelligence agencies expressed alarm that Chinese possession of the system could enable capability replication via “distillation” — a technique where a secondary AI learns by analyzing outputs from an advanced model.

Anthropoc clarified that White House conversations never addressed Chinese penetration concerns. Instead, those discussions centered on Fable 5 and allegations of its security circumvention.

Mythos Capabilities Explained

Introduced in April 2026, Mythos represented a cutting-edge cybersecurity AI deemed too hazardous for public distribution by Anthropic.

The system possesses autonomous capabilities to detect and weaponize previously unknown software weaknesses. Documentation from Anthropic revealed it discovered thousands of critical zero-day exploits spanning major platforms and web browsers.

Among its achievements: uncovering a 27-year-old remote crash exploit in OpenBSD and scoring 83.1% on the CyberGym vulnerability assessment. Anthropic calculated it could reduce expert-level penetration testing timeframes from months to mere minutes.

Distribution was restricted to approximately 12 designated corporate allies and over 40 institutions participating in Project Glasswing, a defensive security initiative.

Unauthorized access incidents occurred previously. A Discord community reportedly maintained illicit Mythos access for fourteen days before detection and termination by Anthropic.

Fable 5 Security Breach and Cryptocurrency Consequences

Fable 5 represents a consumer-accessible variant of Mythos enhanced with protective limitations. White House AI strategist David Sacks revealed via X that Anthropic’s CEO Dario Amodei received notification about the security compromise but minimized its implications.

Anthropoc contested this characterization, asserting the exploit was limited in scope and involved vulnerabilities already existing in publicly available models.

Sacks characterized the circumstance as “serious but easily resolved” while noting Anthropic must take corrective action.

The prohibition significantly impacts cryptocurrency and blockchain security operations. Coinbase had initiated negotiations for Mythos access through Project Glasswing for protective applications.

Prior to restrictions, Zcash’s creator asserted Mythos had comprehensively audited the Zcash blockchain without detecting vulnerabilities.

Security departments dependent on Mythos for weakness identification currently lack alternatives. Anthropic has provided no timeline for service restoration.

The post Anthropic’s Mythos AI Allegedly Compromised by China-Linked Hackers: What You Need to Know appeared first on Blockonomi.