Your conversations, prompts, and creations may be training the next AI model. I analyze the privacy policies that determine what happens to your data when you use AI services.
Side-by-side comparison of privacy scores across major AI services.
| Service | Score | Grade | Key Privacy Issue | |
|---|---|---|---|---|
|
A
Anthropic (Claude)
Best in Category
|
65 | B- | Clearer data practices, transparent training policies | View Review → |
|
P
Perplexity
|
52 | C | Search history profiling, query patterns analyzed | View Review → |
|
O
OpenAI (ChatGPT)
|
48 | C | Training data opt-out complexity, conversation history retained | View Review → |
|
G
GitHub Copilot
|
45 | C- | Code snippets may be retained, telemetry collection | View Review → |
|
G
Google Gemini
|
42 | C- | Integrates with Google account data ecosystem | View Review → |
|
M
Midjourney
|
38 | D+ | All prompts/images public by default on free tier | View Review → |
Most services can use your inputs to train models. Opt-out options exist but are often buried in settings or require explicit requests.
Retained indefinitely unless explicitly deleted. Even after deletion from your account, data may persist in backups and training sets.
Your prompts may pass through multiple services. When AI is embedded in other apps, your data flows through additional third parties.
Even "deleted" data may persist in model weights. Once used for training, your inputs become part of the model itself.
We also analyze AI service terms of service for consumer fairness issues like billing, cancellation, and liability.
View ToS Watchdog AI Services →Click through for complete privacy analysis of each AI service.
If you're worried about how AI services are using your data, or need help understanding your rights under privacy laws, I can help.
Need help with AI data privacy issues?