66% of UK AI Users Spill Their Deepest Secrets to Chatbots, Yet Distrust Tech Companies

A recent survey by Proton, a Swiss technology company, found that 66% of artificial intelligence (AI) users in the UK have discussed at least one sensitive topic with a chatbot. The findings highlight a major paradox in AI adoption: users are increasingly open about sharing personal matters, yet trust in the companies behind the technology remains low.

The survey, which involved respondents in the UK, found that 42% of users have little or no trust in AI companies to protect their personal information. In fact, 11% said they do not trust them at all. Even so, 55% acknowledged that AI has changed the way they make everyday decisions.

The most frequently discussed topics include personal finances (33%), work or career issues (32%), mental health (27%), and romantic relationships (20%). Interestingly, although family and friends remain the top choice for confiding, AI offers the convenience of a loyal, non-judgmental listener available 24/7.

The main concern is not merely whether AI remembers conversations, but how that data could be misused. As many as 64% of UK respondents worry their vulnerable conversations are turned into detailed advertising or marketing profiles. Meanwhile, 52% fear their secrets will become training material for future AI models.

"AI learns far more about us than search engines ever could. We don't just ask questions, we share deep personal context about health, finances, and relationships. When technology knows that much, privacy cannot be a second priority," said Eamonn Maguire, Director of AI Engineering at Proton.

In response to these concerns, Proton launched Lumo, an open-source chatbot that prioritizes privacy with end-to-end encryption, similar to their email service. Lumo is positioned as a safer alternative to ChatGPT. Proton also introduced a tool called AI Paper Trail that visualizes how much personal information leaks from ordinary AI conversations.

The survey also shows that 82% of users would be more likely to use a chatbot designed specifically for privacy, and 44% would be more willing to share sensitive information if there were assurances that their conversations would not be used for AI training.

Indonesian Context

A similar phenomenon could occur in Indonesia as adoption of generative AI rises. Although no similar survey exists yet, awareness of data privacy in the country remains low. Personal Data Protection regulations (UU PDP) are in place, but enforcement is not yet optimal. Indonesian users need to be wary of the risks of sharing sensitive information with chatbots, especially those run by foreign companies. In addition, local technology companies are encouraged to develop transparent and secure AI solutions.

Going forward, the question is no longer whether AI will become more involved in our personal lives, but how much control we have over the data we provide. Will regulators and industry be able to address these concerns before public trust collapses?