Anthropic has released Claude Fable 5, the first publicly available model from its powerful Mythos class of artificial intelligence systems, while imposing safeguards that restrict its use in areas such as cybersecurity and biological research.

The San Francisco-based AI company announced the launch on Tuesday, describing Fable 5 as its most capable generally available model to date and the first broad deployment of technology that had previously been limited to selected partners because of concerns over misuse.

Anthropic said Fable 5 outperformed its earlier models across software engineering, research, image analysis and scientific applications. The company is simultaneously making Claude Mythos 5 available to a restricted group of cybersecurity organisations and infrastructure providers through its Project Glasswing programme, with plans to expand access through a trusted-access scheme.

According to Anthropic, Fable 5 includes automated safeguards that redirect certain cybersecurity, biology, chemistry and AI distillation queries to the less capable Claude Opus 4.8 model. The company said the fallback system was triggered in fewer than 5 per cent of sessions, allowing most users to access the model’s full capabilities.

“Fable 5’s capabilities exceed those of any model we’ve ever made generally available,” Anthropic said in its announcement. The company added that it had adopted conservative safeguards to enable a faster public release while reducing the risk of harmful use.

The launch follows months of restricted testing of Mythos-class models. Anthropic initially limited access because of concerns that the technology could help identify vulnerabilities in critical infrastructure, including banking systems and power grids. The company said it had conducted extensive red-teaming exercises and a bug bounty programme involving more than 1,000 hours of external testing, with no universal method found to bypass its protections.

Anthropic highlighted a range of benchmark results and customer trials to support its claims. Stripe reported that Fable 5 completed a codebase-wide migration in a single day that would otherwise have required a team working for more than two months. Internal research teams said Mythos 5 accelerated parts of the drug-design process by around ten times and generated scientific hypotheses that researchers advanced to experimental evaluation.

The company is pricing both Fable 5 and Mythos 5 at $10 per million input tokens and $50 per million output tokens, less than half the cost of the earlier Mythos Preview programme. Anthropic said Fable 5 would be available immediately through its API and enterprise offerings, while broader access to Mythos 5 would be expanded gradually in consultation with the US government.

Speaking to BBC Newsnight last week, Anthropic co-founder Jack Clark warned that the rapid pace of AI development required stronger mechanisms for restraint. “You want the option to be able to take your foot off the gas and put your foot on the brake,” he said.

---