ANTHROPIC
Anthropic’s “too powerful” model meets the off switch
Anthropic has paused access to its new Claude Fable 5 and Mythos 5 models after US authorities raised national security concerns just days after launch.
The company said it was ordered to block foreign nationals from using the models, but to stay compliant, it had to disable them for all customers.
Not exactly the smoothest product rollout.
Fable 5 is part of Anthropic’s Claude family and was positioned as a powerful rival to the latest models of OpenAI’s ChatGPT and Google’s Gemini.
Anthropic had previously described the model as “too powerful”, especially because of its ability to help with cybersecurity tasks.
The main concern appears to be jailbreaking, which means bypassing a model’s safety limits.
Anthropic said it reviewed the technique and found it was used to identify a small number of already known, minor vulnerabilities.
The company also said other public models could find similar issues without the same bypass.
Three things matter here:
Anthropic says the pause is about complying with a US order, not because it discovered a major new flaw.
The biggest concern is whether advanced AI can help users find or exploit cyber vulnerabilities.
The case could fuel a bigger debate over who gets access to powerful AI models, and who gets to decide.
Big model, bigger headache
The suspension adds to a growing debate over how advanced AI models should be released, tested and controlled.
Some experts warn that limiting access could make it harder to study these systems safely, while others argue that stronger controls are needed as AI becomes more capable in sensitive areas like cybersecurity.
The move also comes as Anthropic is already in a legal dispute with the Trump administration over a separate order affecting government use of its AI tools.
This is why you don’t give your model a villain-origin-story press cycle.- MG