Pew survey: two-thirds of Americans say AI is advancing too quickly

A new wide-ranging Pew Research Center survey lays out Americans' mixed feelings about artificial intelligence. According to The Verge, the survey of 5,000 adults found that a 64% majority sees potential benefits in the technology, but 66% believe it is advancing faster than it can be governed.
The survey asked questions across four main categories: everyday AI use, the impact on work and employment, expectations of government regulation, and trust in the technology. Breaking the results down by age, income, education and political preference revealed interesting patterns.
On everyday use, 41% of respondents said they had used at least one AI chat tool (ChatGPT, Claude, Gemini or similar) in the past six months. That is a marked rise from 28% in 2024. Usage reaches 67% among 18- to 29-year-olds and falls to 18% among those aged 65 and over.
The job front is where the most cautious feelings emerge. Some 58% of respondents think their own occupation will be transformed by AI in the next decade; 32% believe their risk of job loss has grown. The concern is particularly concentrated in middle-class white-collar work, including law and accounting.
Expectations of regulation are high. A 71% majority say the federal government should have more control over AI. Strikingly, both Democratic and Republican voters share this view (78% and 63% respectively). The Trump administration relaxed some AI rules in 2025, but public sentiment runs in the other direction.
Dr Aaron Smith, lead researcher at Pew, told The Verge: "Public opinion is not static; it is shifting rapidly." Smith said the same survey in 2023 returned more positive results, and concern has risen sharply in the past two years. Deepfake videos and election-period disinformation have been particular drivers.
There is a sharp split on trust. The survey measured trust in AI companies at 24%, in the federal government at 31% and in universities at 56%. The relatively strong faith in academic institutions suggests the public cares about which actors are steering scientific progress.
The demographic breakdown is also interesting. The share of those who believe AI improves their life rises to 51% among higher-income groups with postgraduate education, while it drops to 22% among lower-income respondents with high-school education. There is a strong correlation between education, digital literacy and optimism about the technology.
In international perspective, US concern levels are comparable to those in other large economies. The European Commission's 2025 Eurobarometer survey found 68% of EU citizens think AI is advancing too quickly. In Asian economies such as Japan and South Korea, the number is lower and adoption rates are higher.
Pew Research says the results are a signal to technology companies. Smith said: "Firms can no longer maintain the posture of 'we build, the public doesn't get it.' The public wants to know where it is in the process; companies are expected to meet the demand for transparency and accountability." The full report is available on the Pew Research website.
Read next

AI coding agents taught robots to install GPUs and tie cable bindings
New research has shown that AI coding agents can directly supervise robots and train them to install GPUs, tie cable bindings and perform other data-centre maintenance tasks. The study points to a vision of physical automation with less human intervention.

Tens of thousands of Fortinet firewalls allegedly hacked in global corporate breach
Cybercriminals claim to have breached tens of thousands of Fortinet firewalls used by major companies around the world, TechCrunch reports. If verified, the incident would be one of the largest corporate security breaches in recent years.

Useful quantum error correction by 2028? Amazon and QuEra's bold roadmap
Amazon Web Services and the neutral-atom quantum computing company QuEra have released a joint roadmap saying useful quantum error correction could be reached by 2028. The roadmap shortens earlier industry estimates.

Midjourney Medical pivots from cat images to full-body ultrasound scans
Best known as a consumer AI image generator, Midjourney has unveiled Midjourney Medical, a full-body ultrasound scanning system that has secured FDA clearance. The company used more than 18 million anonymised clinical scans to train the model.

Apple's 2027 roadmap rumors: AirPods with cameras for AI and a second folding iPhone
The Verge has aggregated fresh Apple supply-chain rumors pointing to two flagship products for 2027: a new AirPods generation with camera modules for AI and a second-generation folding iPhone.
