Practical guides to protect yourself, your family, and your business from AI-driven scams, deepfakes, and emerging cyber threats.
Popular articles
Memory is on by default
If you have used ChatGPT, Claude, Gemini, or Copilot in the last year, the assistant probably knows things about you that you never explicitly told it during this session. Names of family members. Your job. Cities you have lived in. Health conditions you mentioned once. Project code names.
This is by design. Memory features make assistants more useful — but they also create a data trail most users have never inspected.
What is stored and where
Each major provider stores three categories of data:
- Conversation history — full transcripts of every chat unless you opt out.
- Saved memories — explicit facts the model extracts and persists across conversations (e.g. 'user is a product manager at Acme').
- Training opt-in flag — whether your data can be used to train future models.
The training flag matters most. Conversation history stays in your account; training data, once incorporated, cannot be removed.
Inspect and delete
Direct links to memory and data settings for the major assistants:
- ChatGPT: Settings → Personalization → Memory. Click 'Manage' to see and delete individual memories.
- Claude: Settings → Privacy. Conversations are not used for training by default on consumer plans.
- Gemini: myactivity.google.com/product/gemini. Auto-delete settings hide here too.
- Microsoft Copilot: Settings → Privacy in copilot.microsoft.com.
Practical privacy rules for AI chat
- Never paste anything you would not paste into a Google Doc shared with your employer. Treat chat assistants as logged.
- Turn off training contributions on consumer accounts. The opt-out is usually a single toggle.
- Use separate accounts for work and personal. Cross-context memory is one of the easiest ways for a chatbot to surface something embarrassing or sensitive.
- Audit memory monthly. Take two minutes to read the memory log. You will be surprised what is there.
- For sensitive topics — health, legal, financial — use the temporary chat or incognito mode. These bypass memory.
Where this is heading
Memory will only get more aggressive. Expect agents that browse on your behalf, complete tasks, and accumulate context across weeks and months. The privacy hygiene you build now is the foundation for safely using those agents later.