WORLD - Anthropic has restored global access to its AI model “Claude Fable 5” following discussions with the United States government on cybersecurity safeguards.
The company said the model is now returning with stronger protections and temporary limits on some features.
Stronger Safety Measures Introduced
Anthropic said the updated version includes improved systems designed to detect better and block requests linked to hacking or cyberattacks.
Some requests may be handled differently than before, especially those involving coding or debugging, as the system continues to be adjusted.
Access to Claude Fable 5 and another model, Claude Mythos 5, was suspended earlier in June after new US export control rules took effect on June 12.
These rules required stricter limits on access for certain users. Anthropic said it was unable to reliably verify users’ nationality quickly, so it paused access globally as a precaution.
The restrictions were lifted on June 30, allowing services to begin coming back online.
Gradual Return Across Platforms
Claude Fable 5 is now available again on Anthropic’s main platforms, including Claude.ai, Claude Code, and Claude Cowork.
The company said it will also restore access through major cloud providers such as Amazon Web Services, Google Cloud, and Microsoft Foundry in phases.
However, access is not yet fully back to normal. Some users on paid and enterprise plans will have reduced usage limits for a short period before switching to a new system based on credits.
Anthropic said the decision to strengthen safeguards followed reports from outside researchers that some protections in the model could be bypassed.
In response, the company said it has improved its filtering systems to reduce the chances of harmful requests getting through.
It said the updated system blocks most known attempts, but may sometimes mistakenly block harmless coding or debugging tasks. Users will be informed if their request is blocked, and some queries may be redirected to a less advanced model.
Looking ahead, Anthropic said it is working with major tech companies, including Amazon, Microsoft, and Google, to create shared guidelines for assessing AI security risks.
The goal is to help measure how serious a potential security issue is and how quickly it should be addressed.
The company is also increasing cooperation with US government agencies on AI safety, including sharing information on risks and supporting joint testing of new systems.
For users, Claude Fable 5 is now accessible again worldwide, but with tighter controls, temporary limits, and stricter safety measures designed to reduce misuse while the system continues to evolve.
