Dario Amodei admits AI suffers from a crisis of trust, saying people worry companies or governments are 'cooking up some new way to screw them over'

Anthropic CEO Dario Amodei used a long post on X to argue that the real problem around AI isn’t just fear, it’s distrust. That’s a subtle but important dodge: he pushed back on the idea that he personally is the source of the public’s gloom, while still admitting the mood is bad.

He opened by taking aim at a common argument in AI policy, the one that says regulation either hands power to a few big players or, if you spread the tech widely through open models, keeps everyone honest. Amodei said that’s too neat. Courts and other institutions can spread power, he argued, and Anthropic has backed policies that would slow frontier AI firms down while helping smaller competitors.

But he also said AI naturally concentrates power. Not because of regulation, he wrote, but because of scaling laws: the more resources poured into a model, the better it tends to get. Open-weight models help a bit, he said, but they still leave the biggest advantage with whoever has the most chips and computing power.

He said the right rules could do several things at once: reduce cyber, bio and alignment risks, restrain frontier companies, and still leave room for open-weight models, along with the risks they bring. He also said he supports a FINRA-like body and the Trump administration’s approach to AI testing.

Then came the trust bit. Amodei said the public’s negative view of AI is a serious problem, but blamed a broader failure of confidence in companies, government and the tech industry. A marketing push won’t fix that, he said, and an ad promising AI will cure cancer would probably just sound fake. The fix, in his telling, is actually doing useful things — especially in biology and medicine — and not overpromising before the results exist.