The US government has lifted export controls on Anthropic’s Claude Fable 5 and Claude Mythos 5, allowing the company to restore access to its latest AI models after nearly three weeks of restrictions for non-US citizens.
According to an Anthropic statement, Fable 5 will become available globally from July 1 across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Pro, Max, Team, and select Enterprise users will receive access under 50% of existing usage limits until July 7, after which the model will move to usage-based credits. Anthropic also said it will restore availability on AWS, Google Cloud, and Microsoft Foundry “as quickly as possible.”
Timeline:
- June 9: Anthropic launched Fable 5 for general users and Mythos 5 for a limited group of trusted Project Glasswing cybersecurity partners. Both models share the same underlying system, although Fable 5 shipped with stronger safety guardrails.
- June 12: The US government imposed export controls after learning of an Amazon research report that showed a prompt could bypass some of Fable 5’s safeguards and identify software vulnerabilities. Anthropic consequently suspended access to both models because it could not immediately verify users’ nationality.
- June 26: The government approved restoring Mythos 5 access for a limited set of US organisations while Anthropic continued discussions on expanding availability.
- June 30: The government lifted export controls. Anthropic announced that Fable 5 would return globally from July 1, while cloud platform availability would follow in phases.
Additional safeguards: Anthropic said it worked with the US government, Amazon and other partners to strengthen Fable 5’s cybersecurity protections before restoring access. The company trained a new safety classifier specifically to detect and block the behaviour described in Amazon’s report after concluding that the reported technique did not expose Mythos-level offensive capabilities.
According to Anthropic, the updated classifier blocks the reported bypass technique in more than 99% of cases. When the classifier flags a request, users will receive a notification and the prompt will instead be routed to Claude Opus 4.8. Researchers from the US Department of Commerce’s Center for AI Standards and Innovation (CAISI) also tested the safeguards and agreed they remain exceptionally strong, the company said.
However, Anthropic acknowledged the stricter classifier increases false positives during routine coding and debugging. It said it will continue refining the system to better distinguish legitimate requests from potential misuse while maintaining a layered “defence in depth” approach to cybersecurity.
Anthropic’s proposed consensus framework for AI jailbreaks: The company said the recent export controls exposed the lack of an industry-wide standard for assessing the severity of AI jailbreaks. Consequently, it has begun working with Amazon,…
Source link
Disclaimer
We strive to uphold the highest ethical standards in all of our reporting and coverage. We blogs.grocliq.com want to be transparent with our readers about any potential conflicts of interest that may arise in our work. It’s possible that some of the investors we feature may have connections to other businesses, including competitors or companies we write about. However, we want to assure our readers that this will not have any impact on the integrity or impartiality of our reporting. We are committed to delivering accurate, unbiased news and information to our audience, and we will continue to uphold our ethics and principles in all of our work. Thank you for your trust and support.
Website Upgradation is going on for any glitch kindly connect at [email protected]
