Public opinion of AI is negative, and Anthropic’s CEO says that is not because he keeps warning people about the risks.

In a post on X on August 15, 2026, Dario Amodei wrote:

“I don’t think it is primarily caused by me or any other AI leader warning about AI’s risks. I think it is fundamentally a crisis of trust. I think that ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over.”

He went further, dismissing the idea that better messaging is the fix:

“I don’t think that a glitzy marketing campaign with a positive spin… is the way to win back that trust… The thing that will work is actually curing cancer.”

Who he is

Amodei co-founded Anthropic in 2021, after leaving OpenAI, and has run the company as CEO since. Anthropic makes the Claude models and Claude Code, and Amodei has built much of his public profile around warning about AI safety risks while also arguing AI could deliver major benefits, like faster drug discovery. His posting on X directly, rather than through a company blog, is unusual for him.

What he gets right, and where it’s incomplete

Amodei is right that public distrust of large tech companies did not start with AI, and framing it as a broader crisis of trust rather than a PR failure is a more accurate diagnosis. People who distrust how social platforms handled their data or how tech giants have handled past promises are bringing that same skepticism to AI, regardless of how any one company talks about it.

Where the argument is incomplete: Amodei frames the fix as a binary between marketing spin and delivering results like curing cancer. That skips over a set of trust concerns that are not about outcomes at all, they are about process. Questions about what data models are trained on, whether companies keep the safety commitments they publish, and how AI affects jobs in the near term are not answered by a future cure for cancer. They are answered by transparency about what is happening now. Naming the problem correctly does not mean the proposed fix covers the whole problem.

Why it’s notable

Anthropic has built part of its brand identity on taking AI risk seriously in public, more so than most of its competitors. Amodei publicly rejecting the idea that this positioning itself might be contributing to public anxiety is a real stance, not a neutral observation. It draws a line between “we should be honest about risk” and “our honesty about risk is not the problem,” and asks the public to accept that distinction on his word.

What it means for builders

The same dynamic shows up at a much smaller scale for anyone shipping an AI feature. Confident marketing copy about what your product can do does not build user trust if the product does not reliably deliver on it. Amodei’s own conclusion, that delivered value earns trust and messaging does not, applies just as directly to a startup’s landing page as it does to an industry-wide backlash.

The practical takeaway: if users do not trust what your AI feature does, do not fix that with better copy or a confidence-boosting demo. Show them exactly what the feature does and does not do, and let it earn trust by being reliable in ordinary use. That is a slower path than a strong launch message, but it is the one Amodei himself says actually works.

End of article