PERSON
Onora O'Neill
The Kantian philosopher who taught us that trust is not a feeling but a judgment—and whose three conditions of competence, honesty, and reliability have become the most rigorous framework for deciding whether reliance on an AI is warranted or merely credulous.
Trust, Onora O'Neill insists, is not warmth or openness or the willingness to be vulnerable. It is a reasoned assessment, based on evidence, that a party is competent in the relevant domain, honest in its communications, and reliable over time. This deceptively simple triad—developed across three decades of Kantian moral philosophy and broadcast to the wider world in her landmark 2002 BBC Reith Lectures,
A Question of Trust—has become the sharpest instrument available for cutting through the haze surrounding
large language models. When a lawyer submits fabricated case citations he trusted a machine to supply, the failure is not merely professional; it is the failure O'Neill had always named: the substitution of
credulity—reliance without evaluation—for trust. She has spent her career arguing that the crisis of confidence in modern institutions is not caused by too little trust but by the persistent confusion of the two. AI has made that confusion acutely dangerous,