Anthropic has made its most powerful AI model publicly available, just two months after describing a version of that same model as too dangerous for general use.
On June 9, 2026, the company released Claude Fable 5, the first model from its Mythos line to the public. It runs on the same underlying system as Claude Mythos, which Anthropic introduced in April and restricted to a small group of vetted cybersecurity partners.
From Restricted Access to a Public Release
When Anthropic first announced Mythos in April, it framed the model as a turning point. The company said Mythos could autonomously find and chain together security flaws, including zero-day exploits, across major operating systems and browsers well enough that releasing it without controls would hand attackers a serious advantage.
Access went only to around 40 organizations through a program called Project Glasswing, including Amazon, Microsoft, Apple, and Cisco, all working on defensive cybersecurity.
What Changed with Fable 5
Fable 5 is built on the same foundation as Mythos, but Anthropic added a layer of safety classifiers before opening it up. According to the company, Fable 5 is state-of-the-art on nearly every benchmark it tracks, with the biggest gains showing up on long, complex tasks such as software engineering, scientific research, visual reasoning, and tasks that run with little human supervision.
On some benchmarks, it scores more than 10% higher than Claude Opus 4.8, Anthropic’s previous top model.
How the Safeguards Work
The new system does not simply refuse risky requests. When Fable 5 detects a prompt related to cybersecurity, biology, chemistry, or attempts to extract its capabilities to train a rival model, the request is quietly handed to the older Opus 4.8 model instead, and the user is told a handoff has taken place.
A separate version, Claude Mythos 5, keeps the cybersecurity safeguards switched off and remains limited to vetted cyber defenders and infrastructure operators through Project Glasswing.
A Release with Awkward Timing
Five days before launching Fable 5, Anthropic published a paper When AI Builds Itself, calling for AI labs worldwide to coordinate on slowing down frontier development, warning that systems are approaching a point where they could improve themselves with little human input.
Some users have already pointed out the contrast between the two versions of the model (Fable 5 vs Mythos 5), with comments describing the split as the public receiving the safe version while trusted institutions keep the more capable one.
The release also lands as Anthropic prepares for a possible IPO by October 2026.
