More than four in ten UK adults are deliberately limiting their use of AI. Generative AI, which continues to evolve at a rapid pace, is now embedded in virtually every piece of software and service, becoming an inseparable part of our digital lives. Yet its rapid spread is not necessarily welcomed by everyone.
A report titled "AI: the growing UK pushback," published on July 10, 2026, by King's College London's Digital Futures Institute and Responsible AI UK (RAi UK), reveals a cooling consumer attitude toward this cutting-edge technology as it permeates society. According to a survey conducted by polling firm Deltapoll in June 2026 among 2,055 UK adults, 42% of respondents said they consciously restrict their exposure to AI tools.
The top reason cited for limiting AI use was deep concern over data privacy and security (29%). This was followed by a more conservative motivation—"not wanting to change how they currently do things" (22%). Avoiding AI is not a matter of lacking the skills to use new tools. Rather, it reflects users actively weighing the risks to their own privacy and deliberately choosing not to use AI as a result.
The Generational Paradox: Heavier Users, Greater Caution
One particularly notable finding in this survey data is a paradoxical gap between generations. The data reveals that those who interact most frequently with smartphones and digital services in daily life are, in fact, the most wary of AI.
Gen Z, who should theoretically be the heaviest users of AI, showed a stronger tendency to self-limit their AI use compared to Millennials and Baby Boomers. The classic stereotype that "young people unconditionally embrace new technology and adopt it fastest" has been thoroughly debunked by this survey's results.
Precisely because they are digital natives, this generation is acutely aware of the mechanisms by which platforms collect data and build user profiles behind the scenes. Their wariness about having their written text, search history, and uploaded images absorbed as new training data for black-boxed large language models (LLMs) is by no means an irrational emotional reaction.
Professor Jack Stilgoe of University College London, co-author of the report, noted: "Companies selling technology assume that the people most familiar with it will like it the most, but our findings show that isn't the case."
This is a sobering fact for the industry. The very demographic expected to serve as early adopters driving AI's spread is, contrary to vendors' assumptions, beginning to erect its own barriers. Even as companies pour tens of billions of dollars into developing AI features, if the users who matter most remain wary and refuse to engage, the business model underpinning the current AI boom itself risks collapse.
Worsening Sentiment and a Sense of "No Escape"
Public sentiment toward AI has clearly deteriorated over time. The share of people who believe AI carries more risks than benefits rose from 48% in October 2023 to 52% in June 2026—now surpassing a majority for the first time. Conversely, those who believe the benefits outweigh the risks fell from 38% to 34%. In under three years, the unbridled enthusiasm for the convenience AI promises has cooled, and a more realistic sense of threat has taken hold among the general public.
Underlying this shift is a strong sense of powerlessness over having one's own data and life drawn into AI systems without consent. Seventy percent of respondents said that avoiding exposure to AI is difficult or impossible, even if they wish to do so.
Users can only watch as AI-generated summaries suddenly appear atop search engine results, background data processing changes with OS updates, and social media feeds become saturated with algorithmic content. Frequent revisions to terms of service and data collection being enabled by default are further amplifying users' wariness.
While 63% of people hold positive feelings toward the National Health Service (NHS) and 51% feel positively about wind power, only 29% express favorable sentiment toward AI. This gap starkly illustrates the presence—or absence—of a sense of "controllability" as part of life's infrastructure. Humans instinctively grow wary of black boxes they cannot control.
Until the opacity surrounding what algorithms process personal data and how that data is used is resolved, consumer wariness will not ease. We have reached a point where companies simply issuing statements that "privacy is protected" is no longer enough to restore lost trust.
"Meaningful Consent" and the Industry's Return to Opt-In
These survey findings point to a fundamental shift in the dynamics surrounding AI. Companies have long assumed that by increasing convenience and deeply embedding AI into user interfaces, consumers would naturally come to accept it. But reality has proven otherwise—consumers are beginning to push back against having tools forced upon them while security and privacy practices remain opaque.
Professor Kate Devlin of the Digital Futures Institute argues that addressing this rising anti-AI sentiment requires giving people "meaningful consent"—the ability to choose when AI is applied, how it engages with their data, and how to opt out in critical situations. Rather than a single tap-through consent screen, what's needed is granular transparency that allows users to toggle individual features on and off.
Consumer pushback is also aligning with data protection regulations being strengthened across countries. Legal frameworks like Europe's GDPR (General Data Protection Regulation), which strictly limit the use of user data beyond its original purpose, are becoming one of the biggest obstacles for AI developers. The greater the erosion of user trust, the more regulators can justify stronger intervention.
Things are already in motion. The architecture behind "Apple Intelligence," recently announced by Apple, is a direct response to this anxiety. While centered on on-device processing, it explicitly requires user permission whenever data must be sent to the cloud for more advanced processing. This is a clear attempt to implement "meaningful consent" at the system level.
A model that forces black-boxed data collection as the price of convenience is no longer sustainable. What consumers want is not an all-knowing AI that handles everything hands-off, but an AI where they always remain in control. How the market receives fully opt-in feature implementations like Apple Intelligence will likely shape the de facto standard for AI interfaces going forward.