Anthropic CEO Dario Amodei says the growing resistance to artificial intelligence cannot be explained simply by pessimistic messaging from industry leaders. His diagnosis is more uncomfortable for the companies building the technology: many people do not trust corporations, governments or the technology sector to use powerful systems in their interest.
The argument emerged after investor Gavin Baker criticised Amodei’s warnings about AI risk. Baker said those warnings had contributed to a US backlash, including opposition to data centres, and urged the Anthropic chief to become a more positive advocate for his industry. Amodei rejected the premise that he had presented an unusually negative picture. He described his public writing as balanced between potential benefits and dangers, while accepting that negative public sentiment is a serious problem.
The disagreement is about cause, not concern
Amodei and Baker agree on one point: AI is facing resistance. They differ sharply on why. Baker’s criticism treats industry rhetoric as a significant cause. If executives repeatedly describe catastrophic risks, the public may reasonably question whether the technology should expand rapidly or whether the infrastructure supporting it belongs in their communities.
Amodei instead frames the backlash as one expression of a much older institutional problem. In his view, people already suspect that powerful organisations will find new ways to disadvantage them. AI has become the latest arena in which that suspicion operates. Changing the tone of corporate communication would therefore address the surface of the problem rather than its source.
He also disputes the claim that his own messaging has concentrated on danger. Amodei points to an essay he wrote about the technology’s potential to improve the world, saying he believed the industry had not offered an inspiring enough account of possible benefits. That defence establishes his intended balance; it does not by itself show how different audiences have interpreted the totality of his warnings and promises.
Promises now carry more weight than promotion
Amodei’s strongest concession concerns delivery. He said the most accurate criticism of AI companies, including Anthropic, is that they have not yet fulfilled their large promises to benefit society. That shifts the debate from communication strategy to observable results. Trust cannot be rebuilt solely through brighter forecasts if people do not experience the gains being advertised.
The source does not measure public opinion, identify which promised benefits people consider most important or establish how much specific AI leaders have influenced the backlash. It reports a dispute between an investor and an executive, not a causal study. Amodei’s trust explanation should therefore be read as his interpretation of the political environment rather than a verified account of every source of opposition.
Even so, his response exposes a problem for an industry that often asks the public to accept two messages at once. AI companies present their systems as capable of extraordinary economic and social benefits while also warning that advanced models require safeguards because their risks may be severe. Those positions are not logically incompatible, but they create a demanding credibility test: companies must demonstrate benefits, explain hazards and accept constraints without appearing to use either optimism or fear selectively.
Regulation is not a binary choice
Baker also argued that Amodei had lost the case for AI regulation. Anthropic has supported some rules, including a California proposal imposing transparency requirements on large AI companies. Critics can view regulation as a mechanism that raises compliance costs and strengthens incumbents, allowing the largest developers to consolidate power.
Amodei called it a false choice to suggest that society must either distribute AI widely without regulation or accept concentration created by regulation. He noted that people outside Silicon Valley may instead see rules as a way to constrain corporate power. Anthropic, he said, tries to design proposals that slow frontier developers while helping smaller competitors.
That is Anthropic’s description of its policy intent. The source does not test whether any particular proposal would produce those competitive effects. Rules can be designed to burden large companies more heavily, but their actual consequences depend on thresholds, enforcement and the resources required for compliance. Intent and market outcome are separate questions.
Amodei’s intervention leaves the industry with a harder task than improving its public relations. If opposition reflects a broad deficit of institutional trust, AI companies must show whom their systems benefit, what trade-offs they impose and why their preferred rules distribute power fairly. The argument may continue over whether warnings fuel fear, but Amodei’s own standard is concrete: companies will be judged less by the confidence of their messaging than by whether their promises become visible in people’s lives.
