Anthropic has released Claude Fable 5, a new AI model designed for high-risk tasks and equipped with built-in safeguards. At the same time, the company is keeping its Mythos 5 model off the general market, limiting access to vetted cybersecurity partners.
Safeguards for sensitive work
Claude Fable 5 is the latest addition to Anthropic's lineup, aimed at enterprises that need to push AI into areas where mistakes could be costly. The company said the model includes protections to handle high-risk tasks. It didn't detail what those safeguards are, but the move signals a focus on safety in applications like critical infrastructure, finance, or national security.
The launch gives organizations a tool that can operate in more sensitive environments without requiring custom safety engineering. For Anthropic, it's a bet that customers will pay a premium for a model that's designed from the ground up to avoid catastrophic errors.
Mythos 5 stays behind closed doors
While Fable 5 is available to customers, Mythos 5 takes a different path. That model is restricted to trusted partners in the cybersecurity world. General availability isn't on the table.
The restriction suggests Anthropic views Mythos 5 as too powerful or too risky for broad release. By limiting access to a select group of cyber partners, the company can control how the model is used and monitor for problems before they spread. It's a departure from the typical rollout pattern, where new models eventually reach the public.
The two models serve different purposes. Fable 5 is for high-risk tasks across industries; Mythos 5 is specifically for cybersecurity applications. Both come from Anthropic's research into safer AI, but the company is taking a measured approach with Mythos 5.
Whether Mythos 5 will ever see a wider release is not yet known. For now, only a handful of partners have access.
