June 9 (Reuters) - Anthropic is rolling out a public version of its Mythos AI model, but with guardrails barring its use in risky areas such as cybersecurity, after a preview earlier this year sent shockwaves globally with its ability to find software flaws.
The new Claude Fable 5 is the most powerful model Anthropic has ever made for wider use, the startup said on Tuesday, touting its performance in software engineering and analytics.
Anthropic has so far limited its access to a group of about 200 organizations including the U.S. government under the Glasswing program, after announcing in April that Mythos had uncovered thousands of software vulnerabilities.
Offering its capabilities more widely may allow the $965 billion company to extend the momentum that has powered its valuation above rival OpenAI just as the two startups at the center of the AI industry race to go public.
The company said it had done extensive testing to ensure that users could not manipulate the new model to bypass its guidelines and perform restricted actions.
"Let's say I'm a college student asking the model like help me find cyber vulnerabilities on X package or code. The model would refuse and Fable 5 will fall back to Opus 4.8 for a response," Dianne Penn, Anthropic's head of product management, research and labs, told Reuters.
Fable 5 will be a more expensive model, but it accomplishes tasks with lower token usage, bringing the overall cost per task down, according to early customer feedback, Penn said.
Anthropic also said users who had access to the preview version of Claude Mythos, the version without guardrails, would be able to upgrade to the new Claude Mythos 5.
The company said it planned to expand access over time through a more "systematic trusted-access program."
Pricing on both models is $10 per million input tokens and $50 per million output tokens, the company said.
(Reporting by Zaheer Kachwala in Bengaluru and Deepa Seetharamanin San Francisco; Editing by Anil D'Silva)
