A DuckDuckGo survey reveals a dangerous gap between how people use chatbots and what they understand about privacy. Of 1,944 U.S. adults surveyed, 56% of those who identify as tech enthusiasts admitted sharing secrets with AI that they've hidden from close friends or healthcare professionals.
The data gets more alarming: 53% of respondents didn't know—or weren't sure—that most chatbots use conversations for model training. Only 25% were aware that chat histories can be legally subpoenaed. This means the majority of people confiding sensitive information in AI systems have little to no idea where that data goes or who might eventually access it.
The disconnect highlights a critical need for clearer disclosure and user education. While AI companies have privacy policies, most people don't read them, and the implications of sharing personal information with systems designed to learn from user input remain poorly understood by the average person.