Does your AI train on your data?
Pick the AI your team actually uses. This checker shows what that tool's own consumer terms let it do with what you type, read straight from the policy, reviewed July 2026. No email, no tracking, your choice stays in your browser.
On the consumer tier, read from the tool's own terms:
| Trains on your input by default | |
|---|---|
| How long it is kept | |
| Can a human read it | |
| The catch |
Source: the source policy. Reviewed July 2026; confirm against the live policy before relying on it.
The same questions, answered by Hush
With Hush the questions barely arise, because we keep no conversations or documents. What you type is processed, then gone.
- Is anything kept? No. Conversations are never saved to our servers, so there is no copy to retain, leak, sell or hand over.
- Trained on? No. There is no conversation or document stored to train on, and we never would.
- Can a human read it? No. There is no conversation or document stored to read.
- Whose legal reach? Hardware we own, no US parent, so there is nothing a foreign order could compel.
Sources and verification
Each verdict is read from the tool's own current published consumer terms and from public reporting, on the tier a typical solo professional uses. Read each policy and confirm before relying on this page: OpenAI ChatGPT · Google Gemini · Microsoft Copilot · Anthropic Claude · Perplexity · xAI Grok · Meta AI · DeepSeek (IAPP). Reviewed July 2026. For the full side-by-side, see the AI data teardown.
Frequently asked questions
How do I check if my AI trains on my data?
Look in the tool's data or privacy settings for a training or model-improvement control. This checker summarises the published consumer terms for eight major tools: ChatGPT, Gemini, Copilot, Claude, Perplexity, Grok, Meta AI and DeepSeek. Reviewed July 2026; confirm against the live policy before you rely on it.
Does turning off training make my AI private?
Not on its own. An opt-out still leaves your text stored and, on several tools, readable by staff, and in 2026 some providers weakened or removed their opt-outs. An opt-out you have to keep watching is not confidentiality.
Is any consumer AI safe for confidential client work?
As normally used on the consumer tier, the mainstream tools store your inputs, many train on them, and several sit under non-UK legal reach. The safe pattern is a tool that keeps no conversations or documents and runs on hardware the provider owns, so there is no conversation or document stored to leak, sell or be compelled. That is what Hush is built to be.
Use AI on the work you cannot paste into the tools above.
Try Hush on a real letter or document, the kind you would never put into a tool that trains on it. No card, no account needed.
Start free →Or run the companion check: is your AI within reach of the US CLOUD Act?