AI use rises, but user trust fails to keep up: Thales report
AI adoption is rising, but user trust is not. Many remain cautious about how AI handles their data, especially when it acts on their behalf.
Small moments like signing up, logging in, or resetting a password are shaping how people judge digital services. A new global study from Thales, based on feedback from more than 15,000 consumers, partners, and IT decision-makers, suggests these everyday interactions often determine whether users stay or leave.
Over half of consumers in the study stated they struggled to log into a service at least once in the past year. Many do not wait around when things go wrong. About 68% said they had abandoned a service or switched providers because the process felt too slow or complicated. Some walk away entirely, while others delay using the service or look for another way in.
Many users are willing to accept extra steps if they feel safer. Around 45% prefer stronger security checks, even if it takes longer to sign up. Only 22% favor faster access with fewer protections. Familiar tools help build trust. Nearly seven in ten respondents said features like multi-factor authentication or passkeys made them feel more secure.
At the same time, most users still lack a clear view of how their data is handled. Only 16% said they fully understand how companies collect and use their personal information.
Most IT leaders said their organizations are already using or planning to use AI tools. Yet confidence among users remains low. Just 23% said they trust companies to use AI responsibly with their data. A much larger group expressed concern about AI systems acting on their behalf, especially when those systems make decisions or interact with services without direct input.
Different industries are seeing very different levels of trust. Banking stands apart, with more than half of respondents saying they are comfortable sharing personal data with financial institutions. This puts it well ahead of other sectors. Government services and healthcare follow, but at lower levels.
Beyond that, trust drops quickly. Insurance and education sit in the middle, while sectors like retail, social media, entertainment, and hospitality rank much lower. News media, logistics, and automotive companies fall to the bottom of the list.
Services tied to money, health, or public systems tend to earn more trust, likely because users expect stricter controls. Platforms focused on content, shopping, or media face more skepticism, especially when data use is less clear.
Business partners face a different set of challenges, but the same root issue. Delays in account setup can slow down projects and affect revenue. Only a small share of users receive login details right away, and even fewer get full access on their first attempt.
When official processes take too long, people find workarounds. Many admit to sharing login details with colleagues, even though this creates security risks that are hard to track.
IT teams are aware of these problems, but progress has been uneven. Most leaders said tools like passkeys are important, yet only about half have put them in place. This gap shows how hard it can be to balance user experience with stronger security.
Danny DeVreeze, Vice President of Identity and Access Management at Thales, believes the 2026 Digital Trust Index shows that as AI adoption is accelerating, trust is struggling to keep pace.
“When AI simply helps people work faster, confidence is high. But when AI starts acting autonomously and making decisions or interacting with systems on a user's behalf, people begin asking harder questions about security, control, and accountability,” he said.
The findings suggest that trust is closely tied to how access and data use are handled day to day. When systems are clear and reliable, users are more likely to stay engaged. When they are slow or confusing, people leave, delay, or find unsafe shortcuts. Companies that simplify access, explain data use, and show how AI systems work may have a better chance of keeping user trust over time.