The AI company that said its own tool was too powerful to release to the public has just released it to the public anyway.
Anthropic launched Claude Fable 5 on Tuesday — a version of its Claude Mythos AI that had been restricted to about 150 organisations for preview since April. Back then, the company said Mythos was so intelligent it could hack computer systems and exploit vulnerabilities, so only vetted groups could use it.
Now Fable 5 is generally available. Anthropic says it comes with safeguards and user limitations, but the company isn't hiding its worry. "Releasing a model this capable comes with risks," Anthropic said in a statement. "Fable's capabilities exceed those of any model we've ever made generally available."
"You want the option to be able to take your foot off the gas and put your foot on the brake. Right now, it's like the AI industry has a gas pedal, but it doesn't have a brake pedal."
Canadian Finance Minister François-Philippe Champagne told the BBC in April that the attention on Mythos was warranted partly because "it's the unknown, unknown." US government agencies have also been testing Mythos, despite Anthropic's ongoing lawsuit against the US Department of Defence over its refusal to allow government use of its AI tools.
Anthropic is expected to become a public company soon, with its private valuation nearing $1 trillion (£747 billion). Consistently releasing stronger AI models helps demonstrate its value to investors.
The roughly 150 groups that had preview access to Mythos will now get Claude Mythos 5 — a version without limitations on cybersecurity or biology uses, depending on their specific needs. So far, those groups have reported more than 10,000 critical security flaws in their systems to Anthropic. The company says it'll expand access to Mythos 5 through a "broader trusted access program" soon.
Both Fable and Mythos are essentially the same model but with different safeguards and access levels. They can work "unattended" on human commands for longer periods than any previous Claude models, according to Anthropic.
Co-founder Jack Clark told BBC Newsnight last week that AI capabilities are expanding so rapidly that there should be a way for the public to slow the technology's advancement. "You want the option to be able to take your foot off the gas and put your foot on the brake," he said. "Right now, it's like the AI industry has a gas pedal, but it doesn't have a brake pedal."
Critics have questioned how much of the hype is marketing spin, but the financial and security concerns are real enough that government leaders are paying attention. For now, the public gets Fable — with brakes, but still a very powerful engine.