AI Backlash: Amodei Calls It a Crisis of Trust

Dario Amodei answers Gavin Baker on the AI backlash

Anthropic CEO Dario Amodei pushed back on Gavin Baker in two posts on X on August 15, 2026, rejecting the idea that his warnings about AI risk are what turned the public against it. The first post starts by thanking Baker. Neither post mentions Meta, whose essay on open weights is a separate document.

In the same thread, Baker (@GavinSBaker) had argued that a claim circulating among very serious people in Silicon Valley was easy to believe because it matched Amodei's public messaging. Amodei's second post answers that point. He says he has split his messaging roughly evenly between risks and benefits, with one major essay on each.

Short clips of his interviews, he adds, turn negative as they spread on social media, since negativity gets clicks. He also gives his reason for writing Machines of Loving Grace: the AI industry, in his view, was not offering an inspiring enough picture of what the technology could do for the world.

Why do people hate AI? Amodei sees anti-AI sentiment as distrust

Amodei agrees the public's dim view of AI is a big problem, but he does not think risk warnings from AI leaders are the main cause. In his words, it is fundamentally a crisis of trust. Ordinary people do not trust companies, governments, or the tech industry, and that distrust goes back decades.

People, he says, always suspect that companies like his are cooking up a new way to take advantage of them. AI did not start that suspicion; it is just the latest iteration of it.

Marketing will not fix AI distrust or public trust in AI, Amodei says

A glitzy marketing campaign with a positive spin will not bring trust back, Amodei says; actually curing cancer will. To him, the promise that AI will cure cancer has become more cliche than inspiring, and most people hear it as deceptive.

The criticism he rates as by far the most accurate points at his own side. AI companies, Anthropic included, have not yet delivered on their big promises to benefit the world, he writes. That failure is totally on the companies and on him, and he says it is the point critics should press instead of arguing about messaging and marketing.

Anthropic is ramping up its work in biology and medicine, he says, and he would rather hold off on promises until there is something real to show. Until then, he feels he has to talk honestly about AI's risks. For credibility and trust, he argues, honesty does no worse, and may do better, than ignoring the risks or steering attention away from them.

Open weights and AI concentration of power: where Zuckerberg differs

Amodei sees AI as a technology whose structure pulls power toward a few, with open weights only a partial answer, while Meta CEO Mark Zuckerberg's essay argues that superintelligence should empower everyone. A model has open weights when the lab publishes its trained weights so others can download and run it. In his first post, Amodei calls it a false choice: either regulation hands AI to a chosen few companies and politicians, or AI is spread widely.

Treating regulation as regulatory capture, and so as concentrated power, is too simple, he writes. Many people outside "this bubble" see regulation as a check on corporate power. He does not fully share that view either, saying it is complicated and depends on what the "regulation" actually contains.

What pulls AI toward concentration, he argues, has nothing to do with regulation and more to do with the extreme implications of the scaling laws. Open weights help a bit but are far from enough, since they only shift some of the concentration toward whoever holds the most compute and chips. The right rules of the road, he adds, can leave room for open-weights models and still deal with the specific risks they bring.

Meta posted Zuckerberg's essay, "The Future is for Everyone," to its newsroom on August 10, 2026. The essay frames the choice as superintelligence held by a few institutions or one that empowers everyone, and it argues for individual empowerment, invention, and a balance of power. It calls government policy necessary, warns that limiting capabilities pushes toward centralization, and says open source works against it.

Any proposal that ends in centralized superintelligence deserves skepticism, the essay says, and policy should keep a healthy balance of power that favors people. Amodei's posts give open weights a smaller role, as a partial fix at best.