Anthropic has launched its Claude Fable 5 model for public use with cybersecurity restrictions, as the company expands access to its most powerful artificial intelligence system while imposing guardrails on high-risk applications, according to Reuters.
Anthropic said the new model, based on its Mythos AI system, is designed to deliver stronger performance in software engineering and analytics but will block requests related to cybersecurity exploitation. The company said safeguards were introduced after earlier previews of Mythos reportedly demonstrated the ability to detect software vulnerabilities, raising global concerns.
The rollout marks a broader expansion of access beyond a select group of around 200 organisations, including the United States government under its Glasswing programme. Anthropic said extensive testing had been conducted to prevent users from bypassing safety controls, with restricted queries automatically redirected to earlier models when necessary.
The company added that Claude Fable 5 is its most advanced model for general use, with improved efficiency and lower token usage that reduces overall cost per task, despite higher pricing at US$10 per million input tokens and US$50 per million output tokens.
Anthropic also plans to gradually expand access through a structured trusted user programme as it continues to balance commercial rollout with safety considerations in the rapidly intensifying artificial intelligence race.
Reuters
