CONCEPT
Intelligent Trust
Onora O'Neill's name for the cognitively demanding alternative to both blanket reliance and blanket suspicion—trust extended where the evidence of competence, honesty, and reliability warrants it, withheld where it does not, and revised without hesitation when evidence changes.
Intelligent trust, as Onora O'Neill developed the concept across three decades of Kantian moral philosophy, is the opposite of both credulity and paranoia. Credulity extends reliance without evaluation, accepting whatever is presented because the presentation is convincing. Paranoia withholds reliance regardless of evidence, refusing cooperation even when it is warranted. Intelligent trust—the disciplined, ongoing, evidence-based assessment of whether specific parties meet the three conditions of trustworthiness in specific domains—is the only rational alternative between these failures, and it is also, O'Neill insists, the only posture adequate to the age of
large language models. The question intelligent trust requires is never “Do I trust AI?” but “Does this specific output, for this specific purpose, from this specific system, deployed within this specific institutional context, meet the conditions under which reliance is warranted?” That question must be asked not once but repeatedly, because the model changes, the task changes, the stakes change, and yesterday's warranted reliance may be today's