AI speaks, but answers to no one

For five thousand years, every serious assertion had a person behind it who was accountable for it. AI has just broken that pact: it speaks with the authority of an expert, but answers to no one for what it says. Whom do you trust when nobody signs?

When a doctor says “it’s benign,” they put their license on the line. When a journalist publishes a thesis, they stake their credibility. When a friend says “everything will be fine,” they wager the relationship. For millennia we have calibrated the reliability of an assertion by what it costs the person making it. Every human statement that carries weight is an act of exposure. AI asserts without that exposure. And it does so with a disarming eloquence.

The trap is cognitive. A study published in PNAS shows that people perceive scientists who present AI-simplified summaries as more credible than those who write their own texts, even while considering them less “intelligent” in the academic sense ¹. Our brain equates fluency with truthfulness: we accept a machine’s polished answer more readily than an expert’s nuanced and sometimes tortuous explanation ¹. AI exploits that evolutionary shortcut without fulfilling the implicit promise. It’s an impeccable suit on a mannequin: all the signals of authority, none of the obligations.

And the problem is not theoretical. In Mata v. Avianca (2023), two lawyers submitted fabricated legal citations generated by ChatGPT to a federal judge — nonexistent cases, fictional airlines, invented reasoning — and defended them as real even when the court pointed out they could not be located ². In Moffatt v. Air Canada (2024), a chatbot invented a nonexistent refund policy; the airline claimed the bot was a “separate legal entity” and the tribunal flatly rejected the argument ³. And in Walters v. OpenAI, ChatGPT falsely accused a private citizen of embezzlement in a case he had no part in. Whom do you sue when the slander has no author? ⁴

That legal question reveals the underlying philosophical problem. Some scholars call it “agency without accountability” ⁵: systems that produce the external form of assertion without the internal force of commitment. AI generates fluency without weight. For the first time in history, a statement can have the eloquence of an expert and the accountability of no one.

We have trained these machines to speak like those who know, but not to answer like those who sign. And our brains — wired over millennia to trust whoever speaks with authority — cannot tell the difference. There is no trust without accountability. And until AI can bear the consequences of what it asserts, it will need a human at the front who can. Any other form of autonomy is not progress: it is recklessness.