Anthropic shipped two versions of the same model on September 1, 2026: Fable 5.1 for everyone with tighter, less trigger-happy safeguards, and Mythos 5.1, an uncapped version locked behind vetted cyber and life-sciences access programs.
Read the full story: Claude Fable 5.1 and Mythos 5.1: Anthropic Splits the Safeguards →
Transcript
Anthropic just released two versions of the same AI model. Claude Fable five point one is for everyone, with safety filters tuned to stop over blocking legitimate work. Biology related refusals dropped eighty five percent. Claude Mythos five point one is the other half: the same model, fully uncapped, available only through a government linked verification program for cybersecurity and life sciences work. Anthropic split its safety dial into two models instead of one compromise. They're not alone either. Google shipped its own gated cyber model just one day later. When three competing AI labs land on the same tiered access strategy in the same week, that's not a coincidence. That's the new normal for how powerful AI actually ships.