Privacy in the AI age is not about hiding. It is about informed consent — knowing what information you are exchanging, for what benefit, and whether that trade is worth it. Most people have never consciously made that calculation. This audit helps you make it now, systematically, in under an hour.
Why AI Changes the Privacy Equation
Traditional data collection was passive — websites tracked where you clicked. AI tools are different: they collect what you think. When you type a question into ChatGPT, you are giving OpenAI information about your concerns, your relationships, your health worries, your professional problems. That is categorically more intimate than a click.
The question is not whether to use AI tools — many are genuinely valuable. The question is: which ones, under what terms, with what data, and with what controls?
Step 1 — Inventory Your AI Tools
List every AI tool you currently use. Be thorough. This typically includes:
- Conversational AI: ChatGPT, Claude, Gemini, Copilot
- Writing assistants: Grammarly, Notion AI, Jasper
- Image generators: Midjourney, DALL-E, Adobe Firefly
- Voice assistants: Siri, Alexa, Google Assistant
- Health apps with AI: Apple Health, Fitbit, Whoop
- Email and productivity AI: Gmail Smart Compose, Superhuman, Otter.ai
Most people list 8-15 tools when they actually count. Write them all down.
Step 2 — Check the Training Data Clause
For each tool, find the answer to one question: Does this company use my conversations to train its AI? Here is the current status for major platforms:
| Tool | Trains on your data? | Opt-out available? |
|---|---|---|
| ChatGPT (free) | Yes (by default) | Yes — Settings → Data Controls |
| ChatGPT Plus | Optional | Yes — same setting |
| Claude (Anthropic) | No (conversations not used) | N/A |
| Gemini (Google) | Yes (by default) | Yes — Activity controls |
| Grammarly | Yes | Limited |
Step 3 — Apply the Sensitivity Filter
Not everything you type into AI tools is equally sensitive. Apply this three-tier filter:
- Tier 1 (Safe to share freely): General questions, research topics, creative writing with no personal information, coding help
- Tier 2 (Share carefully): Work documents with company information, general health questions, relationship topics — use tools with clear opt-outs or enterprise data protections
- Tier 3 (Avoid sharing with standard consumer tools): Specific medical diagnoses, financial account details, legal matters with personally identifiable information, anything under NDA
Step 4 — The Practical Adjustments
Based on your audit, make these three changes:
- Opt out of training data on every tool that allows it. (ChatGPT: Settings → Data Controls → toggle off “Improve the model for everyone”)
- Use a separate browser profile for sensitive AI conversations so they are not linked to your main Google/Apple identity
- For work use, verify your company has an enterprise agreement — most enterprise plans explicitly exclude your data from training
The Proportionality Principle
Perfect privacy is not achievable online, and obsessing over it is counterproductive. The goal is proportionality: the sensitivity of what you share should be proportional to the trustworthiness of the tool and the value you receive. A recipe app does not need your location. A navigation app does. Apply the same logic to every AI tool you use.
Key Takeaway: Run the audit once, make three changes, apply the sensitivity filter going forward. Privacy in the AI age is an ongoing practice, not a one-time setting.
Be the first to respond