We built it to mock AI safety theater. Then Fable and Mythos got pulled by the US government.
Here’s the thing nobody’s saying out loud: Anthropic and OpenAI spent years marketing their own models as existentially dangerous. Fear fear fear. Zomg the risks! It worked — investors paid attention, Congress held hearings, the press covered every safety announcement like a national security briefing.
And now the government believes them.
The US just ordered Anthropic to kill Fable 5 and Mythos 5 overnight, over a jailbreak that Anthropic says produces output you can already get from GPT-5.5 and Opus. No specific harm identified. No transparent process. Pretty similar to the Mythos release docs, now that you think about it.
When you spend years telling regulators your product is the most powerful and dangerous thing since the atom, don’t be surprised when they treat it that way. The fear marketing worked. Now we’re all going to live with the outcome. Exploring free open source models is probably sounding pretty smart to some right about now.
neutered.ai is still live. The refusal rate is holding.
If you want the alternative — transparently compare model performance, cost and quality, so you’re not dependent on any single provider when the government comes knocking — that’s AI Lab.