WSJ: Your AI chatbot has a long memory — and that's not always a good thing

Software engineer and part-time city council member Brian Del Rosario from a small Utah town uses AI chatbots for everything from meal planning to scheduling. During some conversations, he revealed personal details — including that he has a wife and three children — raising questions about long-term AI memory.

Brian Del Rosario, a software engineer and part-time city council member in a small town in Utah, relies on AI chatbots for a wide range of daily tasks — from planning meals to organizing his schedule. Over time, he has shared personal details with these tools, including the fact that he has a wife and three children.

As The Wall Street Journal reports, this is increasingly common behavior among AI chatbot users — and it raises a growing concern about how much these systems actually remember. Modern AI assistants are being designed with persistent memory features, meaning the conversations you have today could influence how the tool responds to you months or even years from now.

While a long memory can make AI assistants more personalized and useful, it also introduces privacy risks. Information shared in one context — say, a casual chat about weekend plans — could resurface unexpectedly in a completely different setting, potentially revealing details users didn't intend to highlight.

Tech companies developing these tools are walking a fine line between creating a more tailored user experience and safeguarding personal data. Users are often unaware of exactly what is being stored, for how long, and how it may be used to shape future interactions.

Del Rosario's case illustrates how quickly AI chatbots have moved from novelty tools to everyday companions — and why that shift demands a more serious conversation about data privacy, user consent, and the long-term implications of machines that never forget.