AI agents are quickly becoming part of everyday life, moving beyond answering questions and generating content to carrying out tasks on users’ behalf. From booking holidays and managing diaries to making purchases and accessing personal accounts, the technology promises to make our lives easier than ever before.
However, as AI agents become increasingly autonomous, questions are emerging about how much control people are willing to hand over and whether adequate safeguards are in place.
Research commissioned by cybersecurity PR agency Eskenzi PR and conducted by Censuswide among 2,000 UK adults found that 84% would give AI agents only moderate or low control over their daily lives, or none at all.
More than a third (34%) wouldn’t feel comfortable giving AI agents any control whatsoever, suggesting that while consumers recognise the potential benefits of AI, many remain hesitant about allowing it to act independently.
Security concerns are holding back trust
As AI agents gain greater access to sensitive information and personal accounts, the potential consequences of security failures become increasingly significant.
Eskenzi’s research found that 36% of Brits are concerned about their personal data being hacked or stolen when using AI agents. Almost a third (31%) worry an AI agent could make a mistake, while 30% are concerned about it having access to too much personal information.
Financial security is another major concern, with 28% worried that an AI agent could spend their money or make purchases they never intended to authorise.
Perhaps most surprisingly, one in seven (14%) fear an AI agent could implicate them in something illegal, rising to 21% among 18 to 24-year-olds.
These concerns highlight the security challenges associated with giving autonomous systems access to personal information and the ability to perform actions on behalf of users.
Unlike traditional AI chatbots, which primarily generate responses to user prompts, AI agents may be connected to external applications and granted permissions to interact with accounts and services. This introduces potential risks around compromised credentials, excessive permissions and malicious instructions designed to manipulate an agent’s behaviour.
For cybersecurity professionals, ensuring these systems operate within clearly defined boundaries will be essential to reducing the risks associated with wider adoption.
People want assistance, not complete autonomy
Despite concerns about security, the findings suggest that consumers are not rejecting AI entirely. Instead, they appear more comfortable using it as an assistant than allowing it to make important decisions independently.
More than a quarter (27%) of respondents would allow an AI agent to prepare tasks for them but would want to approve every action. A further 23% would allow AI to handle routine tasks independently while requiring permission for more significant decisions.
There are also clear generational differences. Almost half of over-55s (47%) would not give AI agents any control over their daily lives, compared with 21% of 18 to 24-year-olds.
The findings suggest that organisations developing AI agents need to consider how much autonomy users are comfortable granting, rather than assuming greater independence will always be welcomed.
Introducing meaningful human approval checkpoints, limiting agent permissions and ensuring users can review actions before they are completed could help organisations address some of these concerns.
Who takes responsibility when AI goes wrong?
Beyond security and privacy, the research raises important questions about accountability.
When asked who should bear primary responsibility if an AI agent with access to their accounts was hacked, tricked or carried out an unauthorised action, 40% of respondents pointed to the company that developed the AI.
A further 25% believed responsibility should sit with the company or service the agent was accessing, while only 18% thought the individual using the AI should be primarily responsible.
While these findings reflect public expectations rather than legal liability, they demonstrate the importance of establishing clear accountability as autonomous systems become more widespread.
The implications extend to businesses deploying AI agents in customer-facing and decision-making processes.
When asked how they would react if a company they relied on used an AI agent to advise its board on important decisions, 37% wanted reassurance that a human would always have the final say.
Meanwhile, 20% said they would consider switching provider, suggesting that concerns around AI oversight could also have reputational and commercial consequences.
Building confidence through security and transparency
AI agents have the potential to deliver significant improvements in convenience and productivity, but their growing autonomy creates new responsibilities for developers and organisations deploying the technology.
Clear permission controls, secure authentication, monitoring, audit trails and human oversight can all play an important role in reducing risks and improving accountability.
Equally important is ensuring that users understand what an AI agent can access, what actions it can perform and how they can intervene if something goes wrong.
The research suggests that consumers are willing to embrace the benefits of AI, provided they can retain control over decisions that matter.
As AI agents become more integrated into everyday life, the challenge for the cybersecurity industry will be ensuring that innovation is accompanied by the security, transparency and safeguards needed to earn public trust.
The research was conducted by Censuswide among a nationally representative sample of 2,000 UK adults between 16 and 18 September 2026, on behalf of Eskenzi PR.
