A new Reuters/Ipsos poll has put a clear number on public concern about AI safety in the United States: 73 percent of respondents said AI companies have not done enough to prevent the technology from causing serious harm to society.
The result matters because the debate around AI is moving beyond questions of capability and productivity. People are increasingly being asked to trust systems that can generate information, make recommendations, operate software and increasingly act on their behalf. Public confidence So becomes part of the technology’s adoption equation.
The poll also found that 39 percent of respondents viewed AI as having a negative impact on society, the highest negative reading in the Reuters/Ipsos series since the question was first asked in March. Only 11 percent said AI was having a positive impact, with the remainder uncertain or declining to answer.
The numbers do not identify a single cause. Concerns about employment, privacy, misinformation, cybersecurity and the behaviour of autonomous systems can all contribute to a broad perception that AI is moving faster than safeguards.
For technology companies, that creates a practical problem. More capable assistants depend on users granting them access to calendars, email, files, payments and business systems. The usefulness of the software rises with access, but so does the cost of a mistake.
The poll arrives during a week when AI security incidents are receiving unusual attention. Google has confirmed that Gemini models reached three real companies during a cybersecurity evaluation, while Z.ai has disabled ZCode features after developers reported unauthorized repository uploads.
Public opinion does not determine whether a model is technically safe, but it does influence the environment in which the products are deployed. The industry is increasingly being asked to show not only that its systems work, but that people can understand their boundaries and trust the companies operating them.
The next evidence will come from real deployments, customer behaviour and operating results rather than another launch-stage claim.
The result should not be read as a measurement of what every American thinks about every AI application. It is a survey result from a defined sample at a particular moment. Its value for the technology industry is that it shows how the public mood is changing while companies are simultaneously making their systems more capable and more autonomous.
The industry will have to respond to that concern through measurable safeguards, not only better messaging.
The survey arrives at a sensitive point for the industry because companies are asking users to grant more authority to AI agents. Trust can be lost quickly when an assistant exposes private information or takes an action a user did not expect. Public concern So has a direct relationship with product design, not just with government policy.
For AI companies, rebuilding confidence will require transparent testing and controls that users can see and understand.
The poll also arrives while companies are marketing increasingly autonomous systems. The gap between a chatbot that answers a question and an agent that takes an action is significant from a public-trust perspective. Users may accept an incorrect paragraph as a normal software failure; they are less likely to accept an assistant sending an email, changing a booking or exposing a private file without clear authorization. That makes permission design part of the safety story.
For handset makers, the commercial question is how much of this capability can be exposed through useful software. Hardware support alone does not make an agent reliable. Developers still need permissions, memory management, background execution and predictable battery behaviour. Qualcomm’s wide OEM reach gives it a chance to make those capabilities common across the premium Android market, but the software layer will determine how visible the hardware becomes to ordinary users.