The OECD released its 2026 Trust Survey this week, and buried inside is a finding that ought to reshape how any government department talks about its AI rollouts. Across 33 countries and five accession candidates, people are more optimistic that AI in the public sector will improve service quality and efficiency than they are that it will be fair, transparent, or protective of their personal data. Confidence sits close to four in ten on tailored services and cost reduction. It falls further when the question turns to fairness, human oversight, and the safety of the data citizens have already handed over.
Read the chapter in the OECD report and the shape of the trust problem is unmistakable.
Most people remain sceptical of AI deployed in the public sector.That is the exact reverse of what a public administrator would prefer. In any tax office, in any welfare office, in any subsidy pipeline, citizens already assume speed as their due. The moment a refund lands one day faster, it becomes the new baseline. Trust is not accumulated at the efficiency edge. It is built, or lost, at the edges of fairness, of explainability, and of what happens to the record a citizen was legally required to file.
Which suggests the standard sales pitch for public sector AI is upside down. Enterprise AI has to prove return on investment. Government AI has to prove its accountability spine first, and the return on investment follows. Every deployment should surface its human reviewer, its audit trail, and its grievance route as visible product features, not as buried compliance. Otherwise each new efficiency claim widens the very gap the survey has just measured. Fast was always going to be the easy half. Fair is the one we still have to ship.
#AIGovernance #PublicSectorAI #GovTech #DigitalTrust #OECD #AIinGovernment #Accountability #TaxAdmin
No comments:
Post a Comment