Nicholas Carlini supports controlled Fable 5 release after Mythos concerns
Technology
Nicholas Carlini, a cybersecurity researcher at Anthropic, has done a 180-degree turn on the company's Mythos AI.
He first raised alarms about its risks but now thinks Fable 5, one of Anthropic's latest AI models, should be released with strict controls.
This shift comes as Carlini and his team talk with US officials about keeping these powerful models safe, especially since there are export restrictions in place.
March 2026 Mythos demo raised concerns
Back in March 2026, Carlini showed how Mythos AI could spot software bugs way better than he was, which sounded cool but also risky if misused.
At first, he wanted to delay releasing it. Now he believes that letting people use Fable 5 safely and responsibly could actually help more than hurt, if it's handled right.