Anthropic shipped Fable 5, the model it spent months saying was too dangerous to release, and is warning customers upfront that it "comes with risks."
The company had been holding back a model called Mythos, citing danger serious enough to keep it off the market. Rather than retire it, Anthropic rebranded it as Fable 5, added guardrails, and released it publicly. The caveat about risks is publicly attached to the launch - an unusual move for an industry that tends to frame every release as a stride toward safer AI.
Guardrails filter what a model will do on request. They don't change what it can do. The capabilities that made Mythos too dangerous to ship are presumably still in Fable 5; what changed is Anthropic's judgment that those capabilities are now acceptable with constraints in place. That is a commercial and ethical calculation, not a technical safety breakthrough.
The rename from Mythos to Fable is either coincidental or a deliberate move away from a name carrying internal baggage. The model is out either way.