Anthropic releases Mythos model

(CNN)– Artificial intelligence company Anthropic is releasing a “Mythos-class” model to the public, but with some guardrails.

The model called Claude Fable 5 blocks questions on dangerous topics like cybersecurity and biological research.

It will allow users to ask Mythos questions, something Anthropic itself once considered too dangerous for general release.

Now if users ask about sensitive issues they’ll be redirected to the older opus four-point-eight version of Claude.

Anthropic says Fable is better at remembering thus a better option to do large, complex tasks with fewer instructions.