Anthropic redeploys Claude Fable 5

Anthropic will restore global access to Claude Fable 5 after the US government lifts export controls on the model.

The company said the controls were applied on 12 June to Claude Fable 5 and Claude Mythos 5, requiring access restrictions for foreign nationals inside and outside the United States. Anthropic suspended access to both models for all users because it said it had no reliable way to verify nationality in real time.

Anthropic said the controls were lifted on 30 June. Fable 5 will become available globally from 1 July on the Claude Platform, Claude.ai, Claude Code and Claude Cowork, with access on AWS, Google Cloud and Microsoft Foundry to be restored as quickly as possible.

Access to Mythos 5 has been restored only for a set of US organisations following government approval. Anthropic said Fable 5 and Mythos 5 share the same underlying model, but Fable 5 has stronger safeguards for general use, while Mythos 5 has fewer safeguards and is limited to trusted partners working on defensive cybersecurity.

The export control directive followed a report by Amazon researchers describing a method for bypassing Fable 5 safeguards. Anthropic said the reported behaviour involved identifying software vulnerabilities and, in one case, producing code showing how a vulnerability could be exploited.

The company said its review found that the technique did not expose unique Mythos-level cyber capabilities. It has trained an improved safety classifier to block the behaviour described in the report, and said blocked requests will be redirected to Claude Opus 4.8.

Anthropic also called for a shared industry framework to assess the severity of AI jailbreaks. It said it is working with Amazon, Microsoft, Google and other Glasswing partners on criteria including capability gain, breadth of capability gain, ease of weaponisation and discoverability.

The company said it is expanding cooperation with the US government on frontier AI security, including pre-release evaluation, faster information sharing and joint research on safeguards.

Why does it matter?

The case shows how frontier AI releases are becoming part of national security and export-control policy, especially when models have advanced cybersecurity capabilities. Anthropic’s response also highlights a broader governance gap: governments and companies still lack a shared standard for judging when a jailbreak is minor, serious or urgent enough to justify intervention. The outcome could influence how advanced AI models are tested, released and restricted across borders.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!