Anthropic Releases Public Version of Mythos AI with Cybersecurity Restrictions
Translated from English, summarized and contextualized by DistantNews.
At a glance
- AI company Anthropic has released a public version of its Mythos AI model, named Claude Fable 5, with restrictions on cybersecurity applications.
- The previous preview version of Mythos gained global attention for its ability to identify software flaws.
- Anthropic aims to extend its market momentum with the new model, which is more powerful but includes safety guardrails.
Artificial intelligence startup Anthropic has launched a public version of its advanced AI model, Claude Fable 5, previously known as Mythos. This new iteration comes with significant guardrails, specifically barring its use in high-risk areas like cybersecurity, following concerns raised by its earlier preview version's capability to discover software vulnerabilities.
The company described Claude Fable 5 as its most powerful model yet for broader public use, highlighting its proficiency in software engineering and analytics. Anthropic had previously limited access to a select group of around 200 organizations, including U.S. government agencies under the Glasswing program, after announcing in April that the Mythos model had identified thousands of software weaknesses.
By offering its capabilities more widely, Anthropic seeks to capitalize on its recent momentum, which has seen its valuation surpass that of rival OpenAI. Both companies are key players in the AI industry's race toward public offerings. Anthropic stated that extensive testing has been conducted to ensure users cannot bypass the model's safety guidelines or engage in restricted activities.
Dianne Penn, Anthropic's head of product management, research, and labs, explained that if a user, such as a college student, asks the model to find cyber vulnerabilities in specific code, it would refuse. Instead, Fable 5 would revert to a less restricted model, Opus 4.8, for a response. Penn also noted that while Fable 5 is a more expensive model, its efficiency in completing tasks with lower token usage could reduce the overall cost per task, based on early customer feedback. Users who had access to the preview version without guardrails can upgrade to Claude Mythos 5. Anthropic plans to gradually expand access through a "systematic trusted-access program."
Letโs say Iโm 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.
Originally published by Dawn in English. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.