A large consumer survey suggests AI data consent is often being given even when people are uncomfortable with the request. Usercentrics calls the pattern “resigned consent”: permission given because refusing feels inconvenient, rather than because the person feels confident about sharing their information.
The company’s research, conducted by Sapio Research in March 2026, surveyed 11,000 consumers across the UK, USA, Germany, Spain, Italy, the Netherlands and Sweden. It found that only 7% were fully comfortable giving an AI assistant access to data without conditions. The published materials put the share who would grant access despite discomfort at about 16% to 17%, depending on the summary used.
AI data consent can look stronger than trust really is
That difference between permission and comfort is important because companies often measure access as a binary event. A user accepts or declines. If the user accepts, a product team may reasonably record that as successful consent. The survey suggests the same click can hide very different attitudes.
The full resigned-consent report says an average of 58% of respondents reported some level of discomfort with AI accessing their data across six categories. Discomfort ranged from 53% for work tools to 64% for financial accounts. Only 7% were fully comfortable with unconditional access, while the report gives a global average of 16% for people who would allow access despite being uncomfortable.
The distinction is not a legal judgement about whether a consent process is valid. It is a behavioural finding from a company-commissioned survey. Usercentrics sells privacy and consent-management technology, so its interpretation has a commercial context. Even with that caveat, the numbers expose a practical problem for AI products: a yes can measure compliance with the interface without measuring confidence in the relationship.
The highest stakes can make people stop instead of shrugging
The survey found the strongest discomfort around financial accounts. In that category, 64% reported some level of discomfort and 24% said they would stop using the AI rather than grant the requested access. Work tools produced a different pattern: discomfort was lower at 53%, while 17% said they would grant access despite discomfort.
That makes intuitive sense without proving a universal rule. Repeated, low-friction requests can become part of the rhythm of using a product. A person who needs an assistant to summarise a calendar or work file may be more willing to click through an uncomfortable permission screen than someone facing a request for access to a bank account.
LiveAIWire has previously covered how an AI agent can expose private details about its owner. The new survey approaches the privacy problem from the opposite direction. Instead of asking what an agent can infer or reveal, it asks whether the human giving it access actually feels comfortable doing so.
A reluctant yes could become fragile adoption
For AI companies, this creates a business problem as well as a privacy problem. Products increasingly become useful by connecting to more of a user’s digital life: email, calendars, documents, work systems and financial services. If people authorise those connections reluctantly, adoption can look healthy until a breach, policy change or unexpected feature makes the discomfort visible.
Usercentrics argues that a reluctant yes and an enthusiastic yes should not be treated as equivalent evidence of trust. That conclusion is the company’s interpretation of its own commissioned research, but it is a useful test for product teams. If a permission screen is designed primarily to maximise acceptance, the resulting metric may say little about whether users understood the request or felt able to refuse it.
The same issue appears in wider questions about AI credibility. LiveAIWire recently reported on public trust in AI decision-making. Trust in a system is not the same thing as granting it permission, and a person can continue using a tool while remaining sceptical about how it handles their data.
Better consent may mean accepting more refusals
A more meaningful consent process could therefore produce a lower acceptance rate. Clearer explanations, genuinely equal choices and more granular permissions may encourage some people to say no. That can look worse on a dashboard while producing a user base that understands what it agreed to.
The survey cannot tell us how every AI user behaves, and it does not establish that all reluctant consent is caused by manipulative design. It captures self-reported attitudes across seven markets at one point in time. Cultural differences are substantial: the report says Germany had a much higher rate of resigned consent than the UK or US.
What it does show is that the simple number most services can see, “permission granted”, is not the whole story. As AI assistants request deeper access to the systems people use every day, the quality of that permission may become as important as the quantity. A product that earns fewer but more deliberate yeses could ultimately have a stronger relationship with its users than one built on habitual acceptance.
There is a further design lesson for AI assistants that want access to several services at once. Asking for email, files, calendar and company data in one step may be convenient for onboarding, but it makes it harder for a user to express different levels of comfort. Granular permissions can create more friction, yet they also produce a clearer signal about what the person actually wants the assistant to see. As agents become more capable, that distinction between broad convenience and deliberate access is likely to become more important.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.