Anthropic introduced Claude Fable 5 on Tuesday, offering the general public a version of its powerful Mythos-class AI model while maintaining safety controls that the company said prevent the system from providing dangerous information.
The model will respond to most queries using the Mythos engine, which Anthropic had previously held back from public release because of concerns about its capabilities, according to a report in The Wall Street Journal. If a user asks about a topic the company considers high-risk — such as building a bioweapon or exploiting a software vulnerability — the system falls back to Claude Opus 4.8, an older and less capable model.
“We wanted to be able to provide this level of intelligence for general users in a safe manner,” Dianne Penn, Anthropic’s head of product management, research and labs, told the Journal.
Fable 5 costs twice as much per token as Opus 4.8, but Penn said it uses fewer tokens for certain tasks, making it potentially more cost-efficient for complex jobs. The model also retains memory better, requiring shorter prompts for large, multi-step tasks, she said.
Security researchers and cybersecurity experts have expressed concern that advanced AI models like Mythos are generating an avalanche of bug reports, the Journal reported. Hackers are likely to attempt jailbreaking — tricking the model into bypassing its guardrails — but Anthropic said it has done extensive testing to make that harder to accomplish.
Since Mythos was first announced in April, the Trump administration has re-evaluated its artificial-intelligence policy, according to the Journal. Anthropic has also faced a series of legal and regulatory battles with the administration over the company’s classification as a supply-chain risk and restrictions on its contracts with the Pentagon.
The company previously restricted access to its full Mythos model to about 150 organizations, including Verizon and Microsoft, which are rushing to security-test their systems before hackers gain access to similar technology. Anthropic is upgrading those early-access partners from Mythos Preview to Mythos 5 and plans to broaden access over time through a “more systematic trusted-access program,” the company said.