US restores Claude Mythos 5 access
Following a temporary global shutdown due to government export-control directives on June 12, the US government has authorized Anthropic to redeploy its cybersecurity model, Claude Mythos 5, to select critical infrastructure defense organizations. The general-purpose sister model, Claude Fable 5, remains offline as discussions continue.
The partial restoration of Claude Mythos 5 shows that national security concerns will increasingly restrict the availability of cutting-edge AI, prioritizing critical infrastructure defense over open access. This creates a challenging precedent where the most capable coding and security models are siloed behind government-approved walls.
- –Mythos 5 bypasses safety guardrails that Anthropic built into Fable 5, making it a powerful but double-edged sword for vulnerability discovery and patch generation.
- –The global shutdown on June 12 was prompted by fears of jailbreaks allowing foreign adversaries to exploit the models' cyberoffensive capabilities.
- –Limiting redeployment to over 100 vetted US organizations leaves independent security researchers and the broader developer ecosystem in the dark.
- –No timeline has been established for the return of Claude Fable 5, leaving general API users without access to Anthropic's latest model tier.
DISCOVERED
1h ago
2026-06-27
PUBLISHED
1h ago
2026-06-27
RELEVANCE
AUTHOR
AnthropicAI