Does Telegram train on your data?
Telegram's privacy policy never states whether your content is used to train AI models. This is a documented absence, not an unread document.
Telegram's privacy policy contains no form of the word "train", no "machine learning" and no "artificial intelligence" anywhere in the capture, on 13 Aug and again on 24 Aug. For a messaging service that describes its data handling in some detail, the complete absence of AI-training language is itself the finding: there is no permission to train stated, and no commitment not to. One thing did change on 24 Aug, and it bears on this. The policy used to say "Unlike other services, we don't use your data for ad targeting or other commercial purposes" — a sweeping assurance a reader could reasonably have read as covering model training. Both no-ads sentences survive, but narrowed, and not identically. The principles section now says "We don't use the contents of any of your chats or contact lists to show you ads"; the no-ads section says "we don't use the contents of your private chats or contacts lists for ad targeting or other commercial purposes". The second is the one that reached beyond advertising, and it now reaches only private chats: the contents of public groups and channels have left the "other commercial purposes" assurance, and everything that is not chat content at all — usage metadata, device information — was never inside it. A new section serves ads on the basis of Telegram's "legitimate interest". Nothing was added about training either way. (This entry previously rendered both narrowings as a single phrase, "the contents of your chats or contact lists", which neither sentence says; corrected 1 September 2026.)
- Telegram privacy policy, captured 2026-08-13 (https://telegram.org/privacy)
- train, training, trains, fine-tune, machine learning, artificial intelligence, AI model, large language model
- 0
- 1 Sept 2026
What each plan can refuse
The same questions, asked of every plan we could identify. Every cell comes from a quoted clause or a documented search - see the evidence above.
- Not established
- Not stated
- —
- —
- Not established
How to opt out of Telegram AI training
Plan by plan, from what Telegram's own policy says about refusing. Where it publishes the steps, they're listed here with the limits they carry. Where there's nothing to switch off, or nothing graded, this says so instead of inventing a menu path.
All accounts
Not graded, so no route is pointed at. Telegram's privacy policy contains no form of the word "train", no "machine learning" and no "artificial intelligence" anywhere in the capture, on 13 Aug and again on 24 Aug. For a messaging service that describes its data handling in some detail, the complete absence of AI-training language is itself the finding: there is no permission to train stated, and no commitment not to. One thing did change on 24 Aug, and it bears on this. The policy used to say "Unlike other services, we don't use your data for ad targeting or other commercial purposes" — a sweeping assurance a reader could reasonably have read as covering model training. Both no-ads sentences survive, but narrowed, and not identically. The principles section now says "We don't use the contents of any of your chats or contact lists to show you ads"; the no-ads section says "we don't use the contents of your private chats or contacts lists for ad targeting or other commercial purposes". The second is the one that reached beyond advertising, and it now reaches only private chats: the contents of public groups and channels have left the "other commercial purposes" assurance, and everything that is not chat content at all — usage metadata, device information — was never inside it. A new section serves ads on the basis of Telegram's "legitimate interest". Nothing was added about training either way. (This entry previously rendered both narrowings as a single phrase, "the contents of your chats or contact lists", which neither sentence says; corrected 1 September 2026.)
What this policy has done since we started watching
3 changes detected since 13 Aug 2026. Each one is a diff between two archived copies, and each links to the passage that moved.
Everything we read
Vendor response
We haven't yet written to Telegram about this verdict, and they haven't disputed, clarified or announced a change to any clause quoted above. When either happens it's published here, word for word.
Report a wrong clause