Guides › AI data privacy
Does this AI train on my data?
Pick a tool and your plan. We tell you in plain words whether it uses your chats, files or recordings to train its AI, and exactly how to switch it off. 21 AI tools, read on each official privacy or help page on 2026-10-29.
Choose a tool to see if it trains on your data
Or scroll down: every tool is in the table below.
Policies change. We show what the official page said on 2026-10-29 and link to it. Not legal advice.
What “training on your data” means
- Training means the company uses what you type, upload or record to improve its next AI models. Some also let staff read samples.
- Opting out usually only stops future use. It often doesn't delete what was already collected, and it doesn't delete your chat history.
- Feedback buttons count. On ChatGPT, Claude and Le Chat, a thumbs up or down can let them use that conversation even after you opt out.
- Business plans are usually protected by contract, but check: on some tools your admin decides.
All 21 tools at a glance
21 tools
| How to opt out | Source | |||
|---|---|---|---|---|
| Canva | Unclear — see policy Personal accounts: Canva may use your content to train its AI depending on your Privacy Settings. The pages we read don't say whether that is on or off by default. | Profile picture → Your account → Privacy controls → switch off the AI training options. | Protected by default Canva Business, Teams and Enterprise: content isn't used to improve AI features, and the setting can't be turned on. Canva Education: student content isn't used. | Checked 2026-10-29: official page for Canva (opens in a new tab) · privacy settings steps for Canva (opens in a new tab) · privacy policy for Canva (opens in a new tab) · Check Canva |
| ChatGPT | Trains by default Free, Plus and Pro: OpenAI may use your chats to train its models unless you turn it off. | Settings → Data controls → turn off “Improve the model for everyone”. Or ask “Do not train on my content” in OpenAI's Privacy Portal. Temporary chats aren't used for training. | Protected by default ChatGPT Business, Enterprise and Edu (and the API): not used for training by default. | Checked 2026-10-29: official page for ChatGPT (opens in a new tab) · business page for ChatGPT (opens in a new tab) · data controls FAQ for ChatGPT (opens in a new tab) · Check ChatGPT |
| Claude | Only if you allow it Free, Pro and Max: Anthropic uses your chats and coding sessions for training only if you allow it in your privacy settings (also if you send feedback, or a chat is flagged for safety review). | Settings → Privacy → switch off the model-improvement toggle (“Help improve Claude”). Incognito chats are never used for training. | Protected by default Claude for Work (Team, Enterprise) and the API: inputs and outputs aren't used for training by default. | Checked 2026-10-29: official page for Claude (opens in a new tab) · business page for Claude (opens in a new tab) · privacy setting steps for Claude (opens in a new tab) · how long data is kept for Claude (opens in a new tab) · Check Claude |
| Cursor | Unclear — see policy With Privacy Mode off, Cursor may store and use your code, prompts and editor actions to train its models. Cursor's pages don't say whether Privacy Mode starts on or off for individual plans. | Turn on Privacy Mode in Cursor's settings (free and paid plans). With it on, your code isn't trained on and isn't kept. | Admin decides Enterprise: Privacy Mode is on by default. Teams: an admin can switch it on for the whole team. | Checked 2026-10-29: official page for Cursor (opens in a new tab) · business page for Cursor (opens in a new tab) · security page for Cursor (opens in a new tab) · Check Cursor |
| Descript | Only if you allow it You're asked when you sign up (“Share data with Descript”). Descript's AI models in use today use no user data; only opted-in data goes into research. | App Settings (Ctrl/Cmd + ,) → Profile → “Share data with Descript” → switch it off. | Protected by default Enterprise drives: data sharing is off by default. | Checked 2026-10-29: official page for Descript (opens in a new tab) · Check Descript |
| ElevenLabs | Trains by default Free and paid individual plans: ElevenLabs uses some of the data you give it to improve its audio models for everyone. | Profile icon → Terms and privacy → Data use → turn off “Improve the models for everyone” → Update your choice. | Protected by default Enterprise: not used for training by default. | Checked 2026-10-29: official page for ElevenLabs (opens in a new tab) · Check ElevenLabs |
| Fathom | Trains by default Fathom uses de-identified customer data to improve its own AI models. The AI companies it uses can't train on your data. | Opt out in My Settings (Fathom's page doesn't name the exact switch). | Admin decides Team Edition: an admin can opt out every user at once in Organization Settings. | Checked 2026-10-29: official page for Fathom (opens in a new tab) · Check Fathom |
| Fireflies | Doesn't train by default Meeting audio, video, transcripts and summaries are never used to train AI models, by Fireflies or its vendors. | Nothing to switch off. | Protected by default Same on every plan. | Checked 2026-10-29: official page for Fireflies (opens in a new tab) · Check Fireflies |
| Gamma | Trains by default Free, Plus, Pro and Ultra: your content can be used to train Gamma's AI by default. | Your avatar → Settings → Data Controls → turn off AI training. | Protected by default Business, Enterprise and Teams workspaces: always off, and the setting is locked. | Checked 2026-10-29: official page for Gamma (opens in a new tab) · Check Gamma |
| GitHub Copilot | Trains by default Copilot Free, Pro, Pro+ and Max: since April 24, 2026 GitHub may use your prompts, suggestions and the code Copilot sees to train AI models unless you opt out. | Profile picture → Copilot settings → “Allow GitHub to use my data for AI model training” → Disabled. | Protected by default Copilot Business and Enterprise: not used for AI model training. | Checked 2026-10-29: official page for GitHub Copilot (opens in a new tab) · GitHub's announcement for GitHub Copilot (opens in a new tab) · Check GitHub Copilot |
| Google Gemini | Trains by default Personal Google accounts: while Keep Activity is on, Google uses your Gemini activity to improve its services, including training generative AI models, and human reviewers may read some chats. | Open Gemini Apps Activity → turn off Keep Activity. Or use a temporary chat. Chats are still kept for 72 hours. | Protected by default Work or school Google Workspace accounts: your content isn't used to train generative AI models outside your organization without permission, and isn't read by human reviewers. | Checked 2026-10-29: official page for Google Gemini (opens in a new tab) · business page for Google Gemini (opens in a new tab) · Check Google Gemini |
| Grammarly | Trains by default Free, Premium and single-user Pro: “Product Improvement and Training” is on by default, so your text can be used to train Grammarly's models. | Go to account.grammarly.com/security/privacy → switch off “Product Improvement and Training”. | Admin decides Business, Education, Pro and Enterprise bought through Grammarly's sales team: off by default. Multi-user team accounts bought on the website: on at first; an admin can turn it off for everyone (account.grammarly.com/admin/data_settings). | Checked 2026-10-29: official page for Grammarly (opens in a new tab) · Check Grammarly |
| Le Chat (Mistral) | Trains by default Personal accounts (Mistral now calls the app Vibe): you're not opted out by default, so your chats can be used to train Mistral's models. | Web: Admin panel → Manage → Vibe → Privacy → turn off “Allow your interactions to be used to train our models”. Phone app: Settings → Data & Account Controls → untick “Enable data sharing”. | Admin decides Team: an admin can turn training off for the whole organization. Enterprise: opted out by default. | Checked 2026-10-29: official page for Le Chat (Mistral) (opens in a new tab) · Check Le Chat (Mistral) |
| Lovable | Trains by default Free and Pro: since September 9, 2026 your data may be used to train Lovable's AI models. Data your app's own users enter isn't used. | Account settings → Preferences → AI model training → turn off “Use my Lovable content for model training”. | Protected by default Business and Enterprise workspaces: excluded from training by default; nothing to switch off. | Checked 2026-10-29: official page for Lovable (opens in a new tab) · Check Lovable |
| Microsoft Copilot | Doesn't train by default New Copilot app (from August 18, 2026), personal account: your prompts, responses and files aren't used to train foundation models. | Nothing to switch off in the new app. Still on the older app? Profile icon → your name → Privacy → turn off “Training on conversation activity” and “Training on voice conversations”. | Protected by default Work or school (Entra ID) accounts and Microsoft 365 Copilot: not used to train Copilot's models. | Checked 2026-10-29: official page for Microsoft Copilot (opens in a new tab) · business page for Microsoft Copilot (opens in a new tab) · older app privacy controls for Microsoft Copilot (opens in a new tab) · Check Microsoft Copilot |
| Notion AI | Doesn't train by default Every plan: by default neither Notion nor its AI providers use your workspace content to train models. | Nothing to switch off. | Protected by default Same on every plan. Enterprise gets zero data retention with most AI models. | Checked 2026-10-29: official page for Notion AI (opens in a new tab) · Check Notion AI |
| Otter | Trains by default Otter trains its own speech models on de-identified recordings and transcripts. Its outside AI providers don't train on your data. | Account Settings → Meetings → Feedback and Training → choose “Don't allow sharing”. Otter says future conversations then won't be used for training. | Unclear — see policy Not stated separately for Business or Enterprise on the pages we read. | Checked 2026-10-29: official page for Otter (opens in a new tab) · Feedback and Training setting for Otter (opens in a new tab) · Check Otter |
| Perplexity | Trains by default Free, Pro and Max: “AI Data Retention” is on by default, so your data can be used to improve Perplexity's AI models. | Account settings → Preferences → turn off “AI data retention”. It only applies from then on. | Protected by default Enterprise Pro and Enterprise Max: your data is never used for AI training. | Checked 2026-10-29: official page for Perplexity (opens in a new tab) · Check Perplexity |
| Runway | Trains, no opt-out found Runway's Terms of Use say your inputs and outputs may be used to train and improve its AI models. We found no opt-out setting on its official pages. | No opt-out found. | Unclear — see policy Enterprise has its own contract. The Enterprise help page we read only says outside model providers can't train on Enterprise content. | Checked 2026-10-29: official page for Runway (opens in a new tab) · business page for Runway (opens in a new tab) · Check Runway |
| Synthesia | Doesn't train by default Synthesia doesn't use your inputs or outputs to pre-train its AI. It only fine-tunes (for example a custom avatar) when you ask in writing. | Nothing to switch off. | Protected by default Same for all customers. | Checked 2026-10-29: official page for Synthesia (opens in a new tab) · Check Synthesia |
| Zapier | Doesn't train by default Every plan: Zapier doesn't train on the data that runs through your workflows. It may learn from how workflows are built (the structure, not the contents). | To stop the structure learning too, use the opt-out form linked on Zapier's security page. | Protected by default Enterprise: opted out of the structure learning automatically. | Checked 2026-10-29: official page for Zapier (opens in a new tab) · Check Zapier |
Not included yet
We leave a tool out when its official pages don't say it clearly (read 2026-10-29):
- Midjourney: its privacy policy mentions training data but doesn't plainly say whether your prompts and images train its models, or how to opt out.
- Jasper: its trust page only says your data isn't used to train outside AI models; it doesn't say what Jasper itself does.
Related: Refund and cancel rules · Team plans by team size · Alternatives to popular AI tools · Browse all tools
FAQ
Does ChatGPT train on my data?
On Free, Plus and Pro, OpenAI may use your chats to train its models unless you turn off “Improve the model for everyone” in Settings → Data controls. Temporary chats aren't used for training. ChatGPT Business, Enterprise and Edu aren't used for training by default. Checked 2026-10-29 on OpenAI's help center.
Does Claude train on my chats?
On Free, Pro and Max, Anthropic uses your chats for training only if you allow it in Settings → Privacy (or if you send feedback, or a chat is flagged for safety review). Claude Team, Enterprise and the API aren't used for training by default. Checked 2026-10-29 on Anthropic's privacy center.
Which AI tools don't train on my data by default?
Of the 21 tools we checked on 2026-10-29: Fireflies, Microsoft Copilot, Notion AI, Synthesia, Zapier. Claude and Descript only train if you allow it.
Which AI tools train on my data unless I opt out?
ChatGPT, ElevenLabs, Fathom, Gamma, GitHub Copilot, Google Gemini, Grammarly, Le Chat (Mistral), Lovable, Otter, Perplexity, Runway (on free and personal paid plans). All except Runway show a setting to switch it off. Checked 2026-10-29.
Are team and business plans safer?
Usually. On most tools here, team, business and enterprise plans aren't used for training by default. A few leave it to your admin (Grammarly team accounts bought on the website start with training on), and a few don't say.
Does opting out delete what they already used?
Not necessarily. Opting out stops future use. Perplexity says data collected before you switch it off can't be removed, and OpenAI says turning off training doesn't delete your saved chats. Delete chats separately if you want them gone.