Anthropic Redistributes Claude Fable 5 and Mythos 5 After Export Control Lifted

모델 출시 | Thu Jul 02 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 4 sources

Following the U.S. government's lifting of export controls, Anthropic redistributed Claude Fable 5 and Mythos 5 with new safeguards.

Analysis

[Anthropic] announced the redistribution of Claude Fable 5 and Mythos 5 ^[1][2]

• Globally available on Claude Platform
• Claude.ai
• Claude Code
• and Claude Cowork starting July 1
• Pro/Max/Team and some Enterprise plans include Fable 5 up to 50% of weekly usage until July 7
• AWS
• Google Cloud
• and Microsoft Foundry to be reactivated
• Access to Mythos 5 restored only to U.S. government-approved organizations

[US Department of Commerce] lifted export controls on Fable 5 and Mythos 5 ^[1][2][3][4]

• Withdrew export restrictions imposed over potential jailbreak concerns
• Negotiations led by Commerce Secretary Howard Lutnick
• Center for AI Standards and Innovation reviewed adequacy of safeguards
• Anthropic agreed to cooperate on proactive detection and response to security risks

[Anthropic Safety Classifier] introduced a new safety classifier to block jailbreaks ^[1][2][4]

• Targets specific jailbreak techniques reported by Amazon researchers
• Blocked requests automatically routed to Opus 4.8
• Over 99% success rate in blocking the technique
• Expanded guardrails for cybersecurity and biology-sensitive requests

[Glasswing Industry Consortium] launched development of a joint AI jailbreak evaluation framework ^[1][2]

• Glasswing partners including Amazon
• Microsoft
• and Google participate
• Establishing a common standard for judging jailbreak severity
• Aimed at triaging new findings and communicating with government and industry
• Proposes four categories: attacker capability acquisition
• scope
• ease of weaponization
• and discoverability

[Anthropic-US Government Cooperation] strengthened pre-release testing and information sharing ^[2][3]

• Established a new pre-release testing protocol
• Expanded information sharing and research cooperation with the government
• Mythos 5 released first only to White House-approved organizations
• OpenAI GPT-5.6 also subject to a similar phased rollout

[Global AI Competitive Landscape] noted rise of Asian AI models and residual regulatory risks ^[2][3][4]

• Asian AI models such as Fugu and Tulongfeng approaching Mythos-level performance
• Need to maintain global competitiveness of U.S. AI emerged
• Regulatory burden coinciding with Anthropic's IPO preparation
• Defense Secretary Pete Hegseth's supply chain risk designation remains in effect

Sources

• [1] Redeploying Fable 5Announcements - Anthropic News
• [2] Anthropic’s long-sidelined Fable 5 is greenlit to return - The Verge AI
• [3] Trump drops restrictions on Anthropic’s Mythos and Fable models - TechCrunch AI
• [4] Anthropic Added a New Security Measure to Get Back Into the Trump Administration’s Good Graces - Wired AI