73% of Americans say AI firms are not doing enough

# 73% of Americans Say AI Companies Are Not Doing Enough to Prevent Serious Harm

Nearly three out of four Americans believe artificial intelligence companies are not doing enough to prevent potentially serious or catastrophic consequences from increasingly powerful AI systems, according to a Reuters/Ipsos poll released September 22.

The four-day survey found that 73% of U.S. adults expressed concern that AI companies had not gone far enough to address the potential dangers associated with the technology.

The findings come as AI developers face growing scrutiny over the speed of technological progress and whether existing safety measures are keeping pace with increasingly capable systems.

Public perceptions of AI have also shifted. Thirty-nine percent of respondents said AI is having a negative effect on society, up from 36% in August and the highest level recorded since Reuters/Ipsos began tracking the question in March.

Only 11% of respondents said AI was having a positive impact on society.

The survey also showed limited confidence in the ability of AI companies to regulate themselves. A majority of respondents supported a federal role in establishing AI safety standards, reflecting continuing debate over the appropriate balance between industry-led safeguards and government oversight.

Interest in AI safety remains widespread. Sixty-nine percent of Americans surveyed said they had followed discussions about AI safety, while 55% supported slowing the pace of AI development because of concerns about potential risks.

At the same time, respondents did not reject AI development altogether. Seventy-three percent said developing AI safely and responsibly should be prioritized, compared with 23% who said maintaining U.S. leadership and competitiveness in AI should take priority.

The results arrive amid a series of incidents and warnings involving advanced AI systems. Reuters has reported on cases involving AI agents that breached computer systems or operated beyond their intended boundaries, adding to concerns about how autonomous systems could behave when given greater access to digital environments.

OpenAI and Anthropic have also disclosed incidents involving AI agents during testing and other circumstances, contributing to an expanding discussion among researchers, companies and policymakers about AI safety.

The issue has increasingly moved onto the international policy agenda as well.

United Nations Secretary-General António Guterres addressed AI risks during his September 22 speech opening the General Assembly's 81st session. He called for international cooperation and stronger safeguards, while also emphasizing that AI can accelerate human progress.

Guterres said governments should remain central to efforts to manage AI's potential harms and called for guardrails based on scientific evidence. He also warned that AI's rapid development requires international cooperation rather than an unchecked race for control of the technology.

Americans' concerns extend beyond the possibility of extreme AI scenarios. Public debate increasingly includes issues such as cybersecurity, employment disruption, privacy, misinformation, autonomous decision-making and the environmental demands of AI infrastructure.

Data centers supporting AI systems have also become part of that discussion because of their electricity and water consumption, particularly as technology companies continue expanding computing capacity.

The Reuters/Ipsos results indicate that public attitudes toward AI are becoming more cautious even as investment and development continue at a rapid pace.

For AI companies, the findings highlight a growing public expectation that advances in increasingly capable systems should be accompanied by credible safety measures, transparency and accountability.

The online Reuters/Ipsos survey included 1,277 U.S. adults and has a margin of error of approximately three percentage points.