It’s our ongoing conundrum: Many US adults believe AI will have a negative impact on society. We’re increasingly skeptical of our governments’ ability to rein in the tech’s more dangerous tendencies. But we keep using AI at increasing rates.
A new study from the Pew Research Center published on Wednesday puts this dilemma into numbers. About half of US adults (49%) use chatbots, with nearly a quarter reporting they use AI daily. That’s up 16% from 2024, when just a third of US adults reported using some kind of artificial intelligence tool.
The ways we’re using AI are changing, too. Smart home devices and wearables, like smart watches and rings, are integrating AI into how they work. This gives us more everyday exposure to AI. About a third of US adults say they have a smart speaker, Pew found, and AI features are showing up in some US adults’ smart doorbells (18%), robot vacuums (13%) and even smart thermostats (11%).
ChatGPT reigns over Gemini. Copilot beats out Meta AI, Grok and Claude.
But while we may use AI, we aren’t blind to the risks it poses. More US adults believe AI will have a negative impact on society (40%), slightly up from a similar 2025 Pew report. That’s compared to 31% who now believe AI will have an equally negative and positive impact. Only 16% say it will be positive.
Part of this change from previous years is likely because the AI tools themselves have dramatically changed, too. AI-created images and videos were once easy to spot by their 11-fingered hands and glitches; now, they’re practically indistinguishable from reality. AI slop is all over our social media feeds. Vibe coding tools are technologically eons beyond the simple chatbots that wowed us in 2022.
As AI use grows, so do the opportunities for the tech to do harm. Nearly two-thirds of US adults (63%) believe AI is advancing too quickly, Pew found. In a recent Johns Hopkins University national survey, the majority of US adults said they want to be able to interact with other humans, not AI, in medical care (79%), legal proceedings (76%) and education (74%). Most (75%) want transparency when they’re interacting with AI; nearly three-quarters of US adults want a ban on AI impersonating people’s faces and voices.
Historically, governments have stepped in to prevent some of these more drastic tech catastrophes. But the US government has been hesitant to pass any significant laws around AI. The only significant one is the Take It Down Act, which just went into effect this spring and lets people request AI-altered images of themselves be taken down from social media. Aside from a few sporadic state laws, AI companies are largely free to set their own rules.
Watch this: AI Is Indistinguishable From Reality. How Do We Spot Fake Videos?
03:15
The Trump administration has said that bureaucratic regulations would slow down innovation and prevent the US from beating China in AI development. But recent advancements in AI capabilities have national security advisors proposing a new requirement that all new AI models must pass a government review before they’re released to the general public. Anthropic, which had a very public fight with the Department of Defense over AI, pulled its most recent Fable 5 model after cybersecurity concerns prompted sudden restrictions by the government.
We’ve seen what happens when AI companies fall short. Many families have sued OpenAI, alleging that ChatGPT encouraged their children to harm themselves and ultimately die by suicide. Grok, Elon Musk’s AI, made millions of sexualized AI images of women and children without their consent earlier this year, drawing international outrage, investigations and lawsuits.
So perhaps it’s not overly surprising that 67% of US adults have little to no confidence in the US government to effectively regulate AI, Pew found. That’s slightly up from 2024. About 60% are not confident US companies will develop and use AI tools responsibly. (Only 17% of US adults have faith in the federal government in general, Pew found in 2025.)
You can see in the graph on the left how partisan views have changed over the past two years.
Republicans and Democrats have flip-flopped with each other when it comes to skepticism of federal AI regulation. In 2024, more Republicans than Democrats reported having little to no confidence in the US government to effectively regulate AI. Now, more Democrats are skeptical.
