The Commerce Department has officially lifted export controls on Anthropic’s two advanced AI models–Claude Fable 5 and Claude Mythos 5.
The decision took effect on June 30, and starting July 1, Anthropic is gradually restoring global access to the models. The company had temporarily restricted access to all users because it was technically impossible to filter access by citizenship in real time.
Hot topic: Bitcoin ETF Inflows Hit Worst Month Ever With $4.5B Outflows
The restrictions were imposed on June 12, just three days after the models’ release, over US concerns about possible bypasses of built-in safeguards and the use of the models in cyberattacks.
Contents
How the US Restrictions on Anthropic Started
The export restrictions were triggered by a report from Amazon researchers. They discovered a method to bypass Fable 5’s safety mechanisms, allowing the model to identify software vulnerabilities. US authorities deemed this a potential national security threat and applied export controls to block access outside the US.
Anthropic responded with talks with the administration and implemented additional security measures. The company trained a new classifier that detects and blocks cybersecurity-related queries with more than 99% accuracy. When a suspicious query is detected, it is automatically redirected to the less powerful Claude Opus 4.8.
Anthropic acknowledged that the new filters may produce false positives on routine programming tasks but called it a trade-off for ensuring broad access.
Read more: Top AI Crypto Tokens June 2026 — FET, RNDR, TAO, NEAR, AGIX
Users Will Get Access to Anthropic’s AI Models
Starting July 1, Fable 5 will be available on Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Access via AWS, Google Cloud, and Microsoft Foundry will be restored “as quickly as possible.” Mythos 5 has been open to select US organizations since June 26, and Anthropic is working to expand access to foreign Glasswing participants.
As part of its cooperation with authorities, Anthropic has taken on a set of commitments:
- Providing pre-release access to new models for evaluation
- Promptly sharing information about hacks and abuse
- Conducting joint research and contributing to industry-wide standards
Commerce Secretary Howard Lutnick confirmed that the agency worked closely with Anthropic over the past two weeks to analyze and approve Fable 5.
Additionally, Anthropic, together with Amazon, Microsoft, Google, and other Glasswing partners, has begun developing a unified system for assessing the severity of AI jailbreaks. The proposed framework evaluates threats across four criteria: expansion of attacker capabilities, range of malicious tasks, ease of turning into a real attack, and accessibility to other attackers. The company has also set up a 24/7 leak monitoring team and launched a HackerOne program for reporting potential safety bypasses.
Learn more: What is AI in Crypto? Complete Guide to AI Agents, Tokens & Use Cases
