Users tell AI chatbots secrets that appear in court records
Turns out, a lot of people are telling AI chatbots things they wouldn't share with anyone else, thinking it's all private.
But according to a Washington Post report, that's not really the case. Over the last two years, chatbot transcripts have been cited in the public record in a dozen court cases.
A DuckDuckGo survey found one in three users shared secrets with bots they'd never tell friends or family, and most had no idea their chats could be accessed by others or even subpoenaed.
Many users unaware conversations train models
AI companies might keep your chat data, which means law enforcement or employers (if you're using work tools) could get access.
There's also the risk of leaks, like when Anthropic's Claude chatbot data was exposed in July 2026.
53% of regular users don't know their conversations can help train future AI models, raising questions about how long these chats stick around and who might see them down the line.
Privacy experts say, don't count on full confidentiality unless there are special protections like incognito modes.