Two-thirds of Americans think AI is advancing too quickly

The more telling number isn't the two-thirds who feel uneasy: it's that nearly half of Americans now use AI chatbots regularly, up from a third in 2024. That adoption curve is moving faster than the concern curve, which means the trust deficit is being built on top of active, growing use rather than ignorance of it.

That tension becomes sharper when set against the pace of capability deployment we've been tracking. The self-driving lab story from Radical AI (covered the same day) illustrates exactly the dynamic the Pew respondents are reacting to: AI is no longer just a consumer chatbot but an autonomous agent running 1,200 materials experiments and generating novel compounds at a rate that outpaces institutional oversight. Public discomfort is not irrational when the frontier is moving from text generation into closed-loop physical-world experimentation. The Zhipu GLM-5.2 coverage adds another layer: open-weights models with million-token context windows are now accessible to anyone, compressing the timeline between frontier capability and broad deployment.

Watch whether any major platform (OpenAI, Google, or Meta) announces a formal public-trust or 'responsible deployment' initiative in the next 90 days citing survey data like this. If they do, it signals the industry is treating the sentiment gap as a commercial risk rather than a communications problem.