AI assistants have changed in a fundamental way over the past two years. They no longer just answer your questions in the moment — many of them remember you. Your name, your job, your family situation, your health concerns, your anxieties. All of it, accumulated across conversations and stored indefinitely.
This persistent memory makes AI more useful in some ways. It also means you’ve been building a detailed profile of yourself inside a system owned by a tech company, whether you consented to it deliberately or just clicked through a setup screen.
This guide walks you through how to find what each major AI assistant has stored about you and how to delete it.
Why This Matters More Than You Might Think
AI memory isn’t like a browser cookie. It’s not a tracking pixel. It’s a curated interpretation of you, built from your own words, that the service uses to inform future interactions.
Unlike cookies, which capture behavior passively, AI memory captures intent, emotion, and context. If you told an AI assistant that you were going through a divorce, dealing with a health scare, or struggling with your finances — even once, months ago — that context is potentially still there, shaping how the system responds to you.
There’s also an exposure risk. In April 2026, a researcher demonstrated that Microsoft Recall’s encrypted database could still be extracted under certain conditions despite Microsoft’s security redesign. This isn’t unique to Recall — any system that stores a rich record of your conversations and context is a high-value target.
And when you’ve shared files in AI chats — uploaded a document for analysis, pasted in sensitive content — those inputs may have fed the memory system. Knowing what’s stored and knowing how to remove it is basic digital hygiene.
ChatGPT Memory
What It Stores
ChatGPT’s memory feature, which became available to free users in mid-2025, stores facts about you that are explicitly noted or inferred from your conversations. It does not store full conversation transcripts in the memory — it stores extracted summaries like “User is a freelance designer in Barcelona” or “User has a dairy allergy.”
However, your actual conversation history is stored separately and is also accessible through the interface.
How to View What’s Stored
- Go to chatgpt.com and log in.
- Click your profile icon in the upper-right corner.
- Select Settings.
- Navigate to Personalization → Memory.
- Click Manage memory.
You’ll see a list of stored memory entries. Read through them carefully.
How to Delete It
- Delete individual entries: Click the trash icon next to any entry you want to remove.
- Delete all memories: In the Memory settings screen, look for “Clear all memories” and confirm.
- Turn off memory entirely: In the same Personalization settings, toggle memory off. This stops new memories from being created but does not retroactively delete existing ones unless you also clear them.
Conversation history is managed separately. Go to Settings → Data controls → Delete all chats to remove your conversation history. Note that conversation data may persist in OpenAI’s systems for some period after deletion per their data retention policy.
Gemini and Google’s Personal Context
What It Stores
Google’s Gemini introduced “Personal Context” as part of Google One, and as of 2025 it pulls information from your Google account activity — Gmail, Google Calendar, Search history — in addition to direct conversation context. This means the memory isn’t limited to what you said in Gemini; it’s informed by a broad view of your Google activity.
Personal Context is reportedly on by default for users with a Google account. If you haven’t checked this setting, there’s a good chance it’s been building a profile.
How to View What’s Stored
- Go to gemini.google.com and log in.
- Click your profile icon in the upper-right.
- Select Manage your Google Account.
- Navigate to Data & Privacy → My Activity → Filter by Gemini Apps.
You can view your Gemini conversation history here and see what queries have been logged.
For Personal Context specifically:
- Open Gemini.
- Click Settings (often represented by a gear icon or accessible through your profile).
- Look for Extensions or Personal Context — the exact menu path has changed in updates, so you may need to explore.
How to Delete It
- Turn off Personal Context: In Gemini settings, look for the option to disable Personal Context extensions (Gmail, Calendar, etc.). Disabling these prevents Gemini from pulling from those data sources.
- Delete Gemini activity: In your Google Account under Data & Privacy → My Activity, you can delete Gemini App activity individually or set an auto-delete schedule (3 months, 18 months, or off).
- Pause Gemini Apps Activity: Under My Activity → Gemini Apps Activity, you can pause activity logging. This stops new activity from being saved.
Note: Disabling Personal Context doesn’t delete what Gemini has already inferred from your account history. For a more complete reset, delete the accumulated Gemini activity.
Microsoft Copilot and Recall
The Recall Situation
Microsoft Recall is a Copilot+ PC feature that takes periodic screenshots of your screen, indexes them with AI, and makes them searchable. If you have a Copilot+ PC and Recall was not explicitly turned off, it may have been running.
Recall’s privacy track record has been troubled. The original release was paused after security researchers demonstrated that the database was initially stored unencrypted. Microsoft subsequently redesigned it as opt-in and added encryption and Windows Hello authentication requirements. A researcher reported in April 2026 that vulnerabilities remained; Microsoft disputed the finding.
How to Check and Disable Recall
- Open Settings on your Windows PC.
- Search for Recall or navigate to Privacy & Security → Recall & snapshots.
- If you find this menu, check whether Recall is enabled.
- To delete stored snapshots, look for Delete all snapshots within the Recall settings.
- To turn it off permanently, toggle it off.
Microsoft Copilot Conversation History
For Copilot’s conversational memory (separate from Recall):
- Go to copilot.microsoft.com.
- Navigate to your activity history.
- You can clear conversation history from there.
Claude (Anthropic)
What It Stores
As of late 2026, Claude’s memory behavior depends on the context. In Claude.ai, there is a Projects feature that allows you to add persistent instructions and context that Claude retains across conversations in that project. This is user-controlled and explicit — you choose what to add.
Claude does not, by default, persist memory across separate conversations outside of Projects. Conversation history is stored in your account and accessible through the interface.
How to Review and Delete
- Log in to claude.ai.
- Your conversation history is listed in the left sidebar.
- To delete a conversation, hover over it and look for the delete option.
- For Projects, open the project and review the stored instructions/context. You can edit or delete these.
For data deletion rights, Anthropic’s privacy policy covers your options for requesting deletion of your data — check their current privacy documentation for the most accurate information on retention periods and deletion processes.
General Principles for Auditing AI Memory
Regardless of which services you use, these principles apply:
Treat Shared Documents as Potentially Persistent
If you’ve uploaded a file to an AI chat — a contract, a medical report, a financial statement — that document may have been processed and potentially summarized or indexed. Even if you delete the conversation, information derived from the document might persist in training data, memory, or logs, depending on the service’s policies.
Before uploading sensitive documents to any AI service, check whether the service uses uploaded content for training, how long it retains uploaded files, and whether there’s a way to delete that specific content.
Regularly Purge Conversation History
Many AI services store your conversation history indefinitely unless you actively delete it. Build a habit of reviewing and purging old conversations that contain sensitive information — just as you might clear browser history or delete old emails.
Read the Memory, Not Just the Risk
It’s one thing to know that AI assistants store information about you. It’s another to actually read what they’ve stored. The stored memory entries often reveal surprising interpretations — the service may have inferred things about your health, relationships, or beliefs that you never intended to share directly.
Reading the memory is a useful exercise in understanding what you’ve revealed over the course of many conversations.
Separate Personal and Professional Use
If you use an AI assistant for work, avoid using the same account for deeply personal conversations. Work-related memory entries can mix with personal ones, creating a profile that spans both domains in ways that could be uncomfortable or harmful.
Consider Where You Store Sensitive Information
AI assistants are genuinely useful for many tasks. But they’re not the right tool for storing personal memories, sensitive documents, or information you’d want to keep under your control long-term.
A dedicated private storage service — one that doesn’t process your content to infer context about you, doesn’t use your files for model training, and gives you clear deletion — is a better home for the documents, photos, and records that make up your personal history.
daftei stores your files and memories without training any third-party AI models on your content, without advertising, and without selling your data. When you delete your account, the data goes through permanent, irreversible erasure after a 30-day grace window.
The AI memory audit is not a one-time task. As these services update their features, new memory mechanisms appear. Schedule a review every few months, check for new memory settings, and make sure the profile these systems are building about you is either accurate, minimal, or gone entirely.