A new survey finds nearly one-third of AI users have told a chatbot something they withheld from people they trust — exposing a growing gap between how private AI conversations feel and how private they may actually be.
WHAT’S HAPPENING
A DuckDuckGo-commissioned survey of 1,944 U.S. adults found that 32% of AI users had told a chatbot something they had withheld from a close friend, parent, colleague, doctor or therapist.
Among self-described AI enthusiasts, that number rose to 56%.
Among parents who use AI, 43% said they had shared something with AI that they had never told one of those trusted people.
WHY IT MATTERS
Chatbots are beginning to occupy a strange new position in people’s lives.
They can feel private.
They don’t interrupt.
They don’t judge.
And they are available whenever someone wants to talk.
But that psychological sense of privacy can be very different from actual technical or legal confidentiality.
The survey found 53% of respondents didn’t know or weren’t sure how conversations could be used for AI training, while only 25% knew AI chats could potentially be obtained through legal process.
That creates a growing mismatch:
People may be treating AI like a confidant before fully understanding what happens to the conversation after they confide in it.
WHO BENEFITS
People can use AI to explore sensitive questions they may feel uncomfortable raising with another person.
AI companies also gain deeper insight into how people communicate, what they struggle with and what kinds of assistance they want.
Privacy-focused AI services could benefit as users become more aware of differences in how providers store, use and protect conversations.
WHO LOSES
Users could expose highly personal information without fully understanding a provider’s privacy settings or retention policies.
That risk becomes more significant as people use AI for questions involving health, finances, relationships, work and family.
But not every chatbot operates the same way.
For example, DuckDuckGo says Duck.ai does not use conversations to train its models and anonymizes requests before passing them to model providers.
WHAT HAPPENS NEXT
AI companies may increasingly have to compete on something beyond intelligence:
confidentiality.
Users could begin comparing AI systems not only by which one gives the best answer, but by which one they trust with the question.
And as conversations with AI become more personal, the distinction between feeling private and actually being private will matter much more.
AI became easy to confide in remarkably quickly.
Now the privacy rules have to catch up.
