Skip to content
AI Inspector UtilityBox

Guide · Privacy

How to audit what you've shared with AI

The question isn't whether your AI history contains secrets. It's whether you'll ever find out before someone else does. Here's a practical, local-only audit you can run today.

1. Start from the data you actually have

Export your history from each platform you use — ChatGPT, Claude, Gemini, DeepSeek, or any other AI via the universal parser — then drag the files into AI Data Inspector. Everything below runs on your own CPU.

2. Run the secret scan first

Open the Scanner tab. It flags API keys, tokens, passwords and private keys across every message — grouped by severity. Work high to low:

  • Critical / high: rotate immediately at the provider.
  • Medium: review context — sample keys in docs are common false positives.
  • Low: usually placeholders; skim and dismiss.

3. Read your system prompts & instructions

The Instructions tab surfaces every system prompt and custom instruction in your archive. These often contain corporate context, internal tool names and personal details — worth knowing exactly what you've wired into each assistant.

4. Search for the sensitive words

In the Explorer, search your own high-signal terms — a company name, a project codename, an employee list, your address. Full-text search covers every message across all platforms at once, so you get one audit, not one per app.

5. Keep an export you control

Export a clean Markdown, text or JSON copy of anything you want to keep — locally generated, locally downloaded. Then delete exports you no longer need, and rotate anything the scan flagged. Repeat the audit when your history grows, or quarterly.

Questions

Why should I audit my AI chat history?

People paste keys, credentials, confidential code and personal data into AI chats every day and then forget. An audit finds that exposure before an export, a leak, or a reused device turns it into an incident.

What should I look for first?

Secrets and credentials — API keys, bearer tokens, private keys, passwords. Then system prompts and custom instructions, which are often more revealing than users expect.

Does auditing require uploading my chats anywhere?

No. AI Data Inspector audits entirely in your browser; the scanner, search and reports never leave your machine.

I found a real key. Now what?

Rotate it at the provider immediately (OpenAI, GitHub, Stripe, etc.), then purge it from the copies of your history that you keep.