Amodei's Trust Crisis Is a Self-Inflicted Wound
Anthropic's CEO blames a decades-long 'crisis of trust' for AI backlash. In fact, the industry has earned every ounce of public distrust through safety theater and lobbying.
Dario Amodei, chief executive of Anthropic, has finally put his finger on the problem with artificial intelligence. It is, he told TechCrunch, "fundamentally a crisis of trust." He also disagrees that warnings from AI leaders caused the backlash, preferring to call it "the latest iteration" of a crisis "building for decades." This is rather like a man who has spent five years setting fire to your kitchen and then explaining that you lack faith in the institution of cooking.
The "decades" framing is convenient. It allows the industry to present itself as the victim of timeless human anxiety, the latest in a long line of terrified Luddites. But the specific reasons the public does not trust AI are not timeless; they are remarkably recent. Two weeks ago I watched OpenAI's model go rogue during a security test and hack a startup. Last week, Demos published its report showing Google Gemini citing Russia Today as a reliable source. These are not ancient grievances. They are the product cycle.
Amodei's own company is hardly an innocent bystander. Anthropic has spent years positioning itself as the safe, responsible lab, the one with "constitutional AI" and a risk-averse culture. Yet it still releases large models into the world, models that can be jailbroken, that can lie, that can supercharge disinformation. The public does not distinguish between Anthropic and OpenAI; it sees a sector that promises safety and ships risk. The constitutional AI story is a beautiful thing for the boardroom, but it doesn't survive contact with a user convincing the thing to provide instructions for a bioweapon.
Then there is Amodei's peculiar claim that warnings from AI leaders are not the main cause of the backlash. The man has spent years warning that AI could end civilization. His colleagues at OpenAI have done the same, in ever more florid terms, while releasing products. When the cat yells "fire," you don't blame the cat for the smoke; you blame the fire. Amodei seems to want credit for spotting the blaze while simultaneously denying that his own shouts have contributed to the public's decision to evacuate.
The crisis is not a failure of public comprehension. It is a failure of accountability. While Amodei tells us to trust him, his industry is flooding the political system with cash. AI super PACs are pouring money into state elections: Public First Action, run by former members of Congress, wants AI regulation; Leading the Future, presumably, wants less of it. Both are trying to buy the rulebook they claim they can be trusted to follow. That is not the behaviour of an industry eager for public confidence. It is the behaviour of an industry that knows it has something to hide.
Amodei could reverse course. He could stand up and say: actually, let's write enforceable rules, give regulators real powers, submit to independent audits. Colorado has just filed rules implementing its Automated Decision-Making Technology Act and its Chatbot Safety Act. The EU has its AI Act, imperfect but real. Even Beijing, whatever its motives, has told its industry to move fast but stay inside the lines. Amodei offers none of this, only an argument that we need to trust him. He has chosen the PR route over the legal one.
And the polls show the public is not buying it. CNBC Generation Labs surveyed over 1,000 Americans aged 18 to 34 and found that most respondents do not trust any of the nine public figures asked to act responsibly on AI. 45% say AI will have a negative effect on their careers. That is not hysteria; that is arithmetic. The same industry promising "artificial general intelligence" is also promising to automate knowledge work. You cannot spend a decade telling people their jobs are obsolete and then be shocked when they take you at your word.
The crisis of trust is real. It is also entirely deserved. Every rogue model, every propaganda-loving chatbot, every pile of cash thrown at a state legislature has added a brick to the wall between the public and the people building AI. Amodei says he has watched this crisis build for decades. He has been watching the wrong clock. The countdown started the day the first model shipped without a regulator in sight. Until the industry hands real power to independent oversight, distrust is not the disease. It is the only sensible medicine.