ChatGPT remembers you: what it already knows, and how to erase it

ChatGPT has a “memory” feature that keeps information about you from one conversation to the next: your name, your job, your preferences, your goals — and sometimes details you dropped without thinking, like where you live, a health concern or an opinion. It's convenient, but it builds a persistent profile. You can view it, turn it off and erase it in a few clicks. And above all, you can avoid putting sensitive things in it: what you don't tell it can't be memorized.

What memory keeps (often without you asking)

Memory isn't limited to what you explicitly ask it to keep. According to the official documentation (OpenAI's “Memory FAQ”) and specialized guides, the assistant can save details it deems useful on its own: something mentioned in passing can become a lasting “memory,” reused in your future answers. Over time, this sketches a portrait of you — all the more sensitive because it accumulates quietly.

View, turn off and erase memory

Four actions put you back in control:

1. 1View: Settings → Personalization → “Manage memories” to see what's saved.
2. 2Erase case by case: delete a memory from that list, or ask ChatGPT to “forget” a specific item.
3. 3Turn it all off: disable “Saved memory” and “Reference chat history” so it stops recording.
4. 4Chat without a trace: use temporary chat, which neither reads nor creates memory.

Why it's a privacy issue

A persistent memory is a profile that follows you. It can shape answers, reflect sensitive information you mentioned only once, and outlive the deletion of the conversations that fed it. As long as a memory exists, it's taken into account — including things you'd have preferred to leave in a single discussion.

The rule that comes first: don't deposit sensitive data

Managing memory is good practice, but it's cleanup after the fact. The rule that comes first acts upstream: if sensitive information never enters the conversation, it can't become a memory — nor a profile, nor a risk.

• Review and clean your memories regularly.
• Turn memory off for sensitive topics, or switch to temporary chat.
• Above all: remove identities, health, identifiers and secrets before sending.

That's what ONYRI Sanitize does: the engine detects sensitive data and replaces it with reversible tokens before sending; detection and the mapping stay in your browser, and only anonymized text reaches the AI. Even with memory on, it only has tokens to remember — not your life.