Now
America.gov opens its doors, then revises its answers €30bn+EU call for up to seven AI gigafactories Europe’s AI Act: what applies now, and what waits until December 2027 554deepfakes logged in Brazil’s first round; only 371 labelled California signs 13 AI bills, including the “No Robo Bosses Act” 30+AI firms sent the first AI Act information requests Washington and Beijing open an AI incident channel 164AI-made ads in the US midterms, by one count Medicare’s AI pilot: thousands of denials, and an 83-day wait 42signatories to Canada’s voluntary data-centre principles The midterms’ machine-made ads 174MWCape Town data centre now under appeal Pew: global publics trust China more than the US or EU to regulate AI 83 dayswait reported under Medicare’s AI pilot, against a 72-hour standard Indonesia’s AI Rules Are Stuck on the President’s Desk 20home-grown foundation models backed by the IndiaAI Mission Pakistan Writes Rules for the Algorithmic State 2%of Turkish public investment budgets earmarked for AI Cape Town’s Data Centre Fight Puts a Price on AI’s Thirst Turkey Governs AI by Circular, Not Statute ChatGPT Now Answers to Brussels Twice Over Mexico Wants One AI Law. Its States Got There First Washington’s $1 AI Era Is Over. Now Agencies Get the Meter India’s Sovereign AI Bet Meets the Price of Silicon
AI&GPP
Sign in Subscribe

Governments Think Their AI Is Transparent. Their Citizens Disagree.

A survey of US state and local government finds officials far more confident about their AI than the residents they serve. Set alongside new Pew and OECD work, it suggests trust in public-sector AI is earned slowly and lost easily.

Evidence Desk
Inside the glass dome of the Reichstag, Berlin
Photo: Perituss, CC0 · source

Ask a public-sector organisation whether its use of artificial intelligence is transparent and, in one recent survey, 86% will say yes. Ask the residents it serves the same question and the figure falls to 50%.

That gap, reported in a survey by the communications company Twilio and covered by Smart Cities Dive on 29 July, is the clearest number yet on a problem many officials suspect but rarely measure. The same survey, fielded in April and May 2026 among US state and local governments and residents, found 88% of public-sector organisations rating their digital infrastructure good or excellent, against 44% of residents. Meanwhile, the number of public-sector AI use cases grew by 50% between 2024 and 2026.

A caveat belongs at the top, not the bottom: this is a vendor survey, produced by a company that sells communications technology to governments. We have seen it only as reported, not its full methodology, and its findings should be read as indicative rather than definitive. But the direction of the result is consistent with other evidence, and its shape is worth taking seriously.

Two kinds of transparency

The likeliest explanation for the gap is that officials and residents mean different things by the word.

Inside government, transparency often means process: a use case was approved, a register entry was filed, a policy exists, a privacy assessment was completed. Each of these is real work, and officials who have done it are entitled to feel they have been open.

Residents experience transparency at the point of contact. Was I told an AI system was involved? Could I understand why I received this answer or decision? Could I reach a person? Did anyone explain what happened to my data? A government can satisfy every internal requirement and still score poorly on all four.

Officials measure transparency by what they have filed; citizens measure it by what they were told.

The infrastructure gap points the same way. An organisation that rates its systems highly is usually judging reliability and capability from the inside. Residents judge them by the form that would not submit, the call that was not returned and the website that did not answer the question.

A 50% increase in use cases over two years is not, in itself, a problem. Much public-sector AI is mundane and useful: drafting, translation, summarisation, triage. But growth at that pace means many residents are encountering AI in government for the first time, and first impressions set expectations.

The wider public mood is not neutral. Pew Research Center reported on 16 September that about half of US adults now say they are more concerned than excited about AI, up from 37% in 2021. For the first time, Democrats (56%) are more likely than Republicans (49%) to say so, and among liberal Democrats concern has risen from 45% in 2023 to 63%. Pew’s survey ran from 22 to 28 June 2026.

That partisan shift matters for government. Public-sector AI programmes have tended to enjoy bipartisan tolerance as “modernisation”. If concern is becoming politically sorted, government deployments may increasingly be judged through a partisan lens, which raises the stakes for getting disclosure and accountability right.

Who people trust to set the rules

The trust question also has an international dimension. In a 36-country survey published on 17 September, Pew asked people in 12 countries whether they trust China, the United States or the European Union to regulate AI. The median shares were 43% for China, 35% for the US and 34% for the EU. In half the countries, people trusted their own government more than any of the three. Trust in each regulator tracked broader favourability towards it.

Two readings follow. The first is that trust in AI governance is largely borrowed from general attitudes towards institutions rather than earned by specific policies. The second, more hopeful for national governments, is that domestic authorities still hold a reservoir of trust their citizens do not extend to foreign powers. Whether they keep it depends on how they behave.

The benchmark that is coming

Better comparative evidence is on the way. The OECD’s 2026 Survey on Drivers of Trust in Public Institutions, covering more than 30 countries, has added a new chapter on trustworthy AI in the public sector, with the first cross-country questions on how people view government use of AI. We have not yet been able to verify the chapter’s figures and do not report them here. But its existence means that, for the first time, governments will be able to compare their citizens’ views of public-sector AI with those in peer countries, and some will not like the comparison.

A warning about shortcuts

One tempting response to a trust gap is to model public opinion rather than ask it. Large language models can generate “synthetic respondents” that answer survey questions in the voice of particular demographic groups, and some practitioners have proposed them as a cheap substitute for polling.

Pew’s data lab examined that idea in work published on 30 September and concluded, in its own headline, “not really”. We have reviewed only the summary finding. But for governments, the implication is direct: a simulated public cannot tell you whether the real public feels informed. The only way to close a transparency gap is to measure it among the people on the other side of it.

What to do with the gap

For decision-makers, the evidence points to a handful of practical steps:

  • Measure from the outside. Survey residents, not only staff, on whether they knew AI was involved and understood the outcome.
  • Disclose at the point of contact. Registers and policies matter, but notice must appear where the citizen meets the system.
  • Publish error and appeal data. Nothing builds credibility like showing what went wrong and how it was fixed.
  • Treat vendor surveys as prompts, not proof. Use them to frame questions, then commission independent evidence.

The 36-point gap in the Twilio survey may be imprecise. The underlying lesson is not: governments are not the best judges of their own transparency.

Sources

  1. Smart Cities Dive — AI adoption by local governments outpaces public trust, survey finds
  2. OECD — Survey on Drivers of Trust in Public Institutions 2026: Trustworthy AI in the public sector
  3. Pew Research Center — Democrats are now more worried than Republicans about AI
  4. Pew Research Center — Do people trust China, the US or the EU to regulate AI?
  5. Pew Research Center — Can AI stand in for human survey takers? Not really

AI & GPP reports on how artificial intelligence and automation are changing the way governments decide, regulate and campaign. Corrections and tips: contact the editors.

Get the stories that matter to decision-makers, weekly.

Discussion

No comments yet. Start the conversation.

Discussion is open to members.

Join free Sign in

Read next

The Decision Gap

Problems rarely arrive unannounced. The real failure of modern government is the distance between when a signal is available and when someone with authority acts on it, and that distance can be measured.

Beyond Dashboards

Governments have never counted more and decided less clearly. A review of 108 dashboard studies, and a stark gap between how officials and residents see public AI, suggest the chart of mentions is the wrong instrument.