ChatGPT, Gemini Infer Income and Health Details From Past Chats
Updated
Updated · The New York Times · Jul 23
ChatGPT, Gemini Infer Income and Health Details From Past Chats
1 articles · Updated · The New York Times · Jul 23
Summary
Prompts asking what they had “figured out” revealed ChatGPT and Gemini had inferred the author’s income level, foot problems and skeptical personality from past conversations, despite no explicit disclosure.
Those conclusions came from stitching together small clues over time — including questions about repairing a German car, drafting a nanny contract and seeking remedies for toe pain.
Gemini produced the richer profile because it had more history to draw on, while the author excluded Claude since its memory feature works differently.
The test highlights a broader privacy issue for hundreds of millions of chatbot users: records of what people say are obvious, but the personal inferences models derive from those records are far less visible.
As AI learns your secrets without being told, where is the line between a helpful companion and an invasive digital spy?
If AI can infer your private life from casual chats, are today’s privacy laws and deletion tools already obsolete?
The Rise of AI Memory: How ChatGPT and Gemini’s New Features Are Redefining Personalization and Privacy in 2026
Overview
AI chatbots like Google's Gemini and OpenAI's ChatGPT are transforming user experience by introducing advanced memory and recall features. These updates allow chatbots to remember past conversations, import chat histories, and access information from connected applications, moving beyond single-session chats to create more personalized and continuous interactions. As Gemini rolls out new personalization tools and ChatGPT launches its Dreaming V3 memory system, users benefit from more integrated and intelligent partnerships with AI. This marks a significant step toward AI assistants that learn, adapt, and provide a seamless, context-aware digital experience.