On June 17, 2026, a new Pew Research Center survey, reported by The Verge, set the tone for the next round of US rulemaking: the Pew AI poll shows 63% of Americans think artificial intelligence is advancing too quickly, even as 49% say they use chatbots at least occasionally. ChatGPT usage has doubled since 2023 to 44%, yet only 16% believe AI will have a positive impact on society.
What changed: usage surged, optimism didn’t
The data points to a widening gap between adoption and trust. According to The Verge’s readout of the Pew findings, younger adults lead on use and skepticism at the same time: 66% of Americans aged 18–29 report using chatbots, but 48% in that group say AI will have a negative impact, and only 14% expect a positive one. Older generations report using chatbots less often and are less negative about AI’s societal effects.
That split matters for policy. High usage usually dampens calls to regulate. This time, it may do the opposite. When the most frequent users say they’re uneasy, lawmakers can claim political cover to press for tighter guardrails without fear of throttling a technology people love. The signal is clear enough to fit on a committee slide deck.
Inside the Pew AI poll findings
The Verge attributes the numbers to Pew Research Center’s latest survey of Americans’ AI experiences and expectations. The topline: more people are experimenting with AI tools, yet a strong majority still wants the brakes tapped. The jump from 33% using chatbots in 2024 to 49% in 2026 highlights how quickly the technology has entered daily life. At the same time, only a small minority—16%—foresees a net-positive societal impact, a figure that reinforces sustained skepticism across age groups.
Pew’s broader work on AI public opinion has shown worries cluster around privacy, misinformation, job displacement, and fairness. Readers can browse that research stream on Pew’s Artificial Intelligence topic page, which tracks perceptions as new tools roll out. This latest snapshot suggests familiarity hasn’t yet translated into comfort.
Why this shapes the next wave of rules
Regulators respond to risk perception as much as risk itself. In the US, agencies already have playbooks open. The Federal Trade Commission has warned companies to keep AI marketing claims substantiated and to avoid deceptive deployments in advertising and consumer products; its guidance is public on the FTC’s AI business resources. The National Institute of Standards and Technology’s AI Risk Management Framework gives organizations a shared language for identifying and mitigating harms like bias, safety failures, and security gaps. And the White House laid out a government-wide agenda in October 2023 for “safe, secure, and trustworthy” AI in its Executive Order fact sheet whitehouse.gov.
What the Pew AI poll adds is a fresh mandate: voters are telling Washington that speed without guardrails is unacceptable. Expect to hear the 63% figure quoted in hearings as lawmakers push for clearer transparency rules, testing requirements for high-risk systems, and tougher standards for how models are trained and deployed in sensitive settings like education, hiring, health, and finance.
Europe’s risk-based approach under the AI Act is already becoming the global reference point. Even without copying those statutes, US committees can point to the same public concerns—safety, fairness, traceability—now backed by domestic sentiment data. That makes cross-Atlantic alignment on core expectations more likely, at least around documentation, incident reporting, and provenance signals.
What companies should do before rules arrive
Public skepticism usually lands first as procurement friction, then as regulation. Enterprises buying AI systems will demand proof of safety, reliability, and accountability from vendors even before laws change. The smart response is to meet the market early: publish short, readable model cards; disclose data sources and synthetic data policies where feasible; offer red-team results; and give customers administrative controls that make compliance auditable.
Address the issues people care about most: privacy by default; clear labeling when users interact with automated systems; visible content provenance for generated media; and responsive channels to contest or correct decisions. These aren’t just checklist items. They’re the signals that build a social license to operate when the public mood is cautious.
Companies also have a messaging problem to fix. The Pew numbers suggest heavy users skew negative. That means glossy demos won’t move sentiment on their own. Demonstrating measurable benefits—hours saved, errors reduced, incidents averted—alongside published risk controls will matter more than slogans.
What to watch next in US AI regulation
Two timelines will run in parallel. Agencies will continue to translate the 2023 Executive Order into sector-specific guidance and reporting duties, a process that can reshape procurement and enforcement even without new statutes. Meanwhile, Congress will decide whether to advance targeted bills that codify testing, disclosure, and liability standards for high-risk uses.
In both tracks, expect the Pew AI poll to become shorthand for a political reality: constituents are using these tools and still want guardrails. That combination tends to favor incremental, testable requirements—safety cases for powerful models, content provenance signals to track synthetic media, and audits tied to real-world harms—over sweeping bans. If public sentiment softens as tools prove safe and useful, the regulatory heat could ease. If high-profile failures pile up, the opposite.
The message for builders and policymakers is the same: earn trust on the record. The data says you don’t have it yet. For more on this, see bloomberg.com.