From "AI Will Save Us" to "Shut It Down": How Public Trust Slipped in 12 Months

After two years of warming up to AI, Americans are cooling fast. Trust that businesses will use AI responsibly fell from 31% to 27% in a year, and nearly half of young adults now say AI does more harm than good. Here's what changed, from biased hiring tools to deepfake heists.

The more we use AI, the less we seem to trust it.

For a brief moment around 2024 and 2025, public opinion on AI seemed to be settling into cautious acceptance. People were trying chatbots, companies were adding AI to everything, and the mood was curious rather than hostile. In 2026, that mood has turned.

The numbers

The Bentley University–Gallup Business in Society survey, conducted in May 2026 with 3,270 US adults, tracks trust in AI year by year. Its key findings:

• Only 27% of Americans now trust businesses "a lot" or "some" to use AI responsibly, down from 31% in 2025. Gallup notes this is still above 2023 and 2024 levels, so it's a reversal rather than a free fall.

• 47% of adults aged 18–29 now believe AI does more harm than good, up from 36% in 2025, 27% in 2024 and 30% in 2023. That 11-point jump was the largest of any age group.

Other surveys point the same way. A September NBC News poll found 70% of Americans are more worried than excited about AI, and 70% say it's taking away job opportunities, compared with just 7% who say it's adding them. A Quinnipiac poll in March found 76% trust AI-generated information only rarely or sometimes.

Editor's note on data: You may have seen a "32% to 28%" figure circulating for trust in AI fairness. We couldn't verify it. The closest published figure is Bentley–Gallup's 31% to 27% on responsible use, which is what we use here.

The irony: usage is up

This isn't a story of people rejecting what they don't know. It's the opposite. Gallup finds 70% of Americans now say they're knowledgeable about AI, up from 64% in 2024. NBC finds more than half use AI tools regularly. Familiarity isn't breeding comfort. It's breeding scepticism.

Gartner's famous "hype cycle" describes exactly this pattern: a peak of inflated expectations followed by a "trough of disillusionment" once reality sets in. We may be sliding into that trough now.

What changed? Real-world harms

1. Biased hiring tools. Years ago, Amazon scrapped an experimental recruiting algorithm after discovering it downgraded CVs that included the word "women's," as in "women's chess club captain," because it had learned from a decade of male-dominated hiring. More recently, a US federal court allowed a nationwide age discrimination case, Mobley v. Workday, to proceed as a collective action, testing whether AI screening software can be liable for discrimination. Researcher Joy Buolamwini's landmark "Gender Shades" study found commercial facial analysis systems misclassified darker-skinned women at error rates of up to about 35%, compared with under 1% for lighter-skinned men. When the gatekeepers to jobs and services are biased, trust erodes quickly.

2. Deepfake scams. In 2024, an employee at engineering firm Arup in Hong Kong transferred about $25 million after a video call with what appeared to be the company's CFO and colleagues. All of them were deepfakes. According to the AI Incident Database, more than half of the 346 AI incidents recorded in 2025 involved deepfakes. When you can't trust a face or a voice, you start to distrust the whole technology.

3. Job anxiety. Young adults are the most worried, and they're also the ones entering a job market where entry-level roles in writing, coding, customer service and administration are the most exposed to automation. Among 18–29-year-olds, 73% told NBC that AI is costing job opportunities.

4. Who controls it. A CNBC Generation Lab poll found majorities distrust every major AI company leader it asked about. People aren't just asking whether AI works. They're asking whose interests it works for.

Trust versus trustworthiness

Philosopher Onora O'Neill makes a helpful distinction: the goal shouldn't be more trust, it should be well-placed trust. Asking the public to "just trust" AI is the wrong ask. The right one is for AI systems and the companies behind them to become demonstrably trustworthy, through transparency, independent audits, accountability when things go wrong, and clear rules.

The public seems to agree. NBC found 81% of Americans say the government isn't doing enough to regulate AI.

Is "shut it down" the answer?

Let's be objective here. The same surveys show people see real benefits: in NBC's poll, 55% believe AI will do more good than harm in scientific research. AI is helping detect diseases earlier, speeding up drug discovery, and improving accessibility for people with disabilities. Most people don't want AI banned. They want it governed.

The falling trust numbers are best read as a warning, not a verdict. The industry still has time to earn back confidence, but only if it stops treating public concern as a marketing problem.

Where in your life have you started quietly double-checking what AI tells you, and what would it take for you to stop?

Sources: Bentley University–Gallup (July 2026 and Dec 2025 releases); NBC News Decision Desk poll (Sept 2026); Quinnipiac via TechCrunch (March 2026); Euronews/CNBC Generation Lab (Aug 2026); Deep Heuristics citing the AI Incident Database; Reuters (Amazon, 2018); Buolamwini & Gebru, "Gender Shades" (2018).

Follow Femonomic for grounded, data-driven takes on technology and the people it affects.