You wouldn’t hand a stranger your diary. But every day, millions of people are doing the digital equivalent — sharing their health worries, relationship struggles, financial anxieties, and work secrets with AI chatbots.
Here’s the catch: by default, most of those conversations are being used to train the next generation of AI models.
What’s Actually Happening
OpenAI’s ChatGPT allows consumer-account conversations to be used for model improvement unless you disable the training settings. Anthropic’s Claude and xAI’s Grok can do the same. Google recently announced it will use photos and other material people upload through Search to train its AI systems by default. And Meta? It has been caught using interactions with Meta AI to personalize content and ads across its services — with reports that contractors reviewing conversations could directly access highly sensitive personal identifying information.
The scale is staggering. A single prompt about travel plans seems harmless. But stack up months of conversations covering your job, relationships, health, and finances, and a surprisingly complete picture of your life emerges.
The New Privacy Tool You Should Know About
Proton recently launched AI Paper Trail, a free tool that shows you exactly what ChatGPT and Claude know about you. You export your conversation history, upload it, and the tool generates a report with:
- An AI Exposure Score based on how many personal data points were extracted
- A breakdown of what the AI knows about your identity, habits, relationships, and interests
- An estimated dollar value of what your data could be worth to an AI provider
The tool currently supports ChatGPT and Claude, with more platforms coming soon. Proton says uploaded data is deleted immediately after analysis.
How to Opt Out — Service by Service
| Service | Where to Find It |
|---|---|
| ChatGPT | Settings → Data Controls → Toggle off “Improve the model for everyone” |
| Google Gemini | Gemini Apps Activity settings → Turn off activity storage |
| Claude | Settings → Privacy → Toggle off “Improve Claude for everyone” |
| Microsoft Copilot | Settings → Privacy → Turn off Training on conversation activity |
| Grok (X) | Settings & Privacy → Privacy & Safety → Grok & third-party collaborators → Untick training box |
One important note: For ChatGPT, turning this off only prevents future conversations from being used. Your past data may already be part of the model.
Bottom Line
The AI industry is racing to collect as much data as possible — your chats, photos, voice recordings, even which apps you use. The good news? Most major platforms now offer opt-out options. The bad news? They’re not always easy to find, and they’re rarely turned on by default.
Take five minutes today. Check your settings. Run an AI Paper Trail audit. Your digital diary is worth protecting.
Tags:
Technology