Americans Trust AI With Almost Nothing

Despite the AI industry making up an ever-growing part of the U.S. and global economy, trust in the systems is still slim. According to the latest polls, Americans remain deeply skeptical of AI, seeing clear benefits only in scientific research and health care and only trusting AI-generated information for recipes and cooking instructions.

The latest poll to ask about AI issues comes from NBC News Decision Desk, taken from Aug. 20-Sept. 1. In the opening question about AI, respondents were asked whether they were more excited or more worried about AI. Only 30% said they were more excited about it, while 70% were more worried. Only 8% said that the risks of AI for society were low or very low, while 67% said they were high or very high.

Those concerns also extend to how Americans expect AI to affect major institutions of society. The poll found that the only topics where people thought AI would do more good than harm were scientific research (55% vs 26%) and health care (47% vs. 30%), while on the issues of K-12 education, the economy, news and information, elections and democracy, higher education, national security, arts and entertainment, and jobs and the workplace, more thought it would do more harm than good.

When it comes to government intervention in AI, most Americans do not think Washington is doing enough. Only 19% said the government is doing enough to regulate AI, while 81% said it is not.

So far, Congress has enacted relatively little legislation specifically regulating artificial intelligence. Most of the Trump administration’s AI executive orders have instead directed federal agencies or shaped federal procurement and regulatory policy, rather than imposing direct nationwide requirements on private AI companies. The federal rules that do directly restrict private uses of AI remain relatively narrow, including laws and regulations targeting nonconsensual deepfakes, AI-generated robocalls, fake reviews, and the export of advanced AI chips and certain model technology.

Neither party had much more trust than the other from the public on making “good policy decisions about AI.” Twenty percent said they trust the Democratic Party more, and 16% said they trusted the Republican Party more. The plurality, 44%, however, said they trusted neither party.

For personal use, recipes, and cooking instructions were the only areas where a majority of respondents said they trusted AI-generated information, at 74%. Only 35% said they trusted AI for information on historical events, 33% for health conditions and medical treatments, and only 13% for parenting or relationship advice.

One of the primary arguments against those who want to regulate AI is that if America doesn’t continue advancing its AI systems, China will take the lead. Yet this concern, that China will become the global leader in artificial intelligence, is only a significant concern for 38%, while the other 62% said China becoming the global leader in AI was somewhat of a concern, not very concerning, or not at all concerning.

Regarding educational uses of AI, despite the technology becoming more ubiquitous, the poll found that Americans are increasingly skeptical of using AI in education. In a summer 2025 NBC poll, 53% said they thought that a school that integrates AI tools into the classroom would better prepare students for the future. In the latest poll, only 44% said the same.

Still, only a minority, 31%, said that no AI use is acceptable in completing school assignments. More said that AI use is acceptable for proofreading and editing, 49%; research and brainstorming, 41%; and organizing or outlining ideas, 38%. Very few (7%) thought writing or generating content for school assignments was acceptable.