50% of Americans: “Hard NO to Government AI Surveillance”

In May 2026, ZeroClick Labs fielded a survey to gauge the sentiment toward a hypothetical AI system, entirely built, trained, and operated by the U.S. government. The survey revealed an interesting discovery: Americans are not outright opposed to the idea of government AI – provided certain conditions are met.

However, when it comes to any form of surveillance, that’s where the discussion ends. Or rather, doesn’t even begin. Whether the benefit is stipulated as proven, whether the oversight is layered with judicial checks, whether the data is anonymized – it doesn’t matter. Whichever way the question is asked, the answer is always the same: NO!

What would Americans be willing to hand off to government AI?

The answer is: just the basics – and even that reluctantly. In fact, when asked which types of data they’d be comfortable with government AI using, the largest single segment (~37%) selected “none of the above.” That’s right, nearly a third of Americans find no data category acceptable for any government AI use.

The distribution hits the ceiling with basic identity information (name, age, address), topping out at ~36%. After that, it collapses rapidly and dramatically, with access to criminal/court records sitting at ~18%, online activity / social media data at ~17%, and the absolute floor – biometric data, landing at barely ~13%.

The use-case ladder (discussed at length in the original analysis) tells the exact same story: of all the government AI applications we tested, only chatbots cleared majority support (51%), and even then, only for answering the basic questions about government services.

Monitoring online activity for national security threats landed at the 4th spot with ~27%, and using government AI to support law enforcement or crime prevention only cleared 24%, ending dead-last on the list (not accounting for 7% “not sure” cohort).

Finally, if we observe general comfort levels, it’s immediately clear that the resistance is real: 52% of respondents expressed discomfort with the government AI system using personal data. While the intensity varies (37% very, 15% somewhat), the pattern is consistent:

If a government AI could reduce violent crime with proven success, but required monitoring all citizens' online activity, would you support it?

That was our question, verbatim - and only 16% said yes. Furthermore, barely 24% said they would accept it, but only under strict, independent judicial oversight. The remaining 50% rejected it outright - ~23% on the grounds that the privacy cost would be too high regardless of benefits, and 27% out of the mistrust towards the government’s crime statistics.

These results aren’t exclusive to our survey - several external studies almost perfectly align with our findings:

• A 2026 ITIF / Center for Data Innovation study found that 54% of Americans believe that AI-powered mass surveillance is “too dangerous and violates privacy and civil liberties,” while only 30% believe it is a necessary trade-off for safety.
  • 46% believe the government should be allowed to use AI surveillance on specific targets with a court-issued warrant.
• A 2026 Stanford HAI annual report shows that the trust in the U.S. government to regulate AI responsibly is at an all-time low of 31%.
• A 2026 Brookings Institution report highlights that public opposition isn’t driven by privacy concerns alone, but by historical skepticism of how “proven success” is measured by the U.S. law enforcement agencies.

Note that the above three studies are only the most recent ones; the approval of the U.S. government to use and regulate AI responsibly is low, but with surveillance in the picture, it’s in the gutter: the independent judicial oversight is the only frame that would prevent outright rejection. Even adding in the segment that supports it unconditionally, the maximum coalition is ~40% - still failing to clear majority support.

Why Americans won’t even negotiate government AI surveillance

Privacy rules, audits, penalties, opt-outs, transparency, due process - every facet of the conditional-trust framework that our respondents named pertains to service AI (e.g., chatbot). For surveillance, they don’t come close to clearing majority support.

Again, these results do not exist in a vacuum. Recent Public Citizen’s synthesis of US AI polling shows 80% of the U.S. adults supporting prioritization of AI safety and data-security rules - even if it slows AI development.

The same report highlights that 72% of Americans want independent experts to safety-test AI products and services, and the same percentage wants the AI industry to be more regulated overall. This is no longer about how the AI works - it’s about what it’s allowed to touch.

And the trade-off most surveillance frameworks rest on - privacy for safety - is the idea 50% of Americans in this dataset are not only not willing to entertain, but outright reject, even when safety benefit is stipulated as proven.

Jordan Parkes, the CEO and founder of ZeroClick Labs, has been building and scaling digital marketing strategies since 2012, leveraging performance-driven SEO and data-based digital marketing solutions to guide the growth of hundreds of companies across the U.S. and Europe.