Does TikTok use your videos to train AI? What its 2026 terms say
The Morse teamPublished 8 min read
Yes, TikTok does use your videos to train AI. As of November 3, 2026, TikTok's US terms of service give TikTok USDS Joint Venture an irrevocable license to "Your Content" that covers "training, testing, and improving our machine learning models and algorithms," and that content includes your videos, your messages, and what you type into TikTok's AI tools. TikTok's US pages name no setting that turns this off. In the European Economic Area, the UK, and Switzerland, you can object, and TikTok weighs each objection case by case.
We read TikTok's US Terms of Service and US Privacy Policy (both last updated July 15, 2026), its EEA and UK Terms of Service and Privacy Policy, its right-to-object page, its help page on downloading your data, and two TikTok newsroom posts on November 3, 2026. We compared them with the Internet Archive's copies of the US terms dated November 2023 and January 22, 2026, and the US privacy policy dated August 19, 2024. Coverage in January 2026 split between "TikTok can now train on everything" and "nothing changed." Both are partly right, and the table below shows which part.
TikTok AI training at a glance
Read on November 3, 2026.
| What you hand TikTok | Inside the US training license? | EEA, UK, and Switzerland | After you delete your account | Source |
|---|---|---|---|---|
| Videos and photos, public or private | Yes, as Your Content | Used, unless you object and TikTok agrees | US: the license survives. EEA and UK: it ends, with two exceptions | US terms, EEA terms |
| Drafts you never post | Collected before you save or post, for captions, effects, and sound picks | Same objection route | No period given | US privacy policy |
| Direct messages | Yes, Your Content says "including messages" | Scanned for safety where the law permits | Same as videos | US terms, EEA privacy policy |
| Prompts to TikTok's AI tools, and the replies | Yes, as Input and Output | Same objection route | Same as videos | US terms, right to object |
Does TikTok use your videos to train AI?
Yes, in two documents at once. The US privacy policy lists, among its uses of your information, to "train, test, and improve technology, such as machine learning models and algorithms." The US terms then make it part of the license you grant by posting. That license lets TikTok "access, reproduce (e.g. to copy), distribute, share, download, adapt or make derivative works" of your content for "operating, improving, and providing the Platform and developing new technologies (including training, testing, and improving our machine learning models and algorithms)."
The license is "non-exclusive, irrevocable, and royalty-free," and "assignable and sub-licensable, including through multiple tiers." The terms also say that, "as between you and TikTok USDS Joint Venture, you own Your Content." So TikTok doesn't own your videos, but it can keep using them for training under a license you can't take back.
Neither document names the models, or says whether "machine learning" includes generative AI or only the recommendation system behind the For You feed. TikTok's AI Alive post from May 13, 2025, which launched its photo-to-video tool, also says nothing about where its training data came from.
Does TikTok train AI on your DMs and private videos?
Nothing in TikTok's US terms or privacy policy excludes them. The US terms define Your Content as "content you create, import, upload, publish, or generate with the Platform, including messages." The definition adds "prompts, questions, files, and other types of information that you submit to our AI-powered interfaces," together with the responses.
The US privacy policy puts training and message scanning in one sentence. The purpose that covers training machine learning models ends with "scanning, analyzing, and reviewing user content, messages, AI interactions, and associated metadata." The policy also says TikTok "may collect user content through pre-uploading at the time of creation, import, or upload, regardless of whether you choose to save or publish that user content."
Instagram draws the line differently. Meta's pages say it trains on public posts, which we cover in does Instagram use your photos to train AI. TikTok's US documents draw no public-versus-private line for training at all. A private account limits which users see your videos, but not the license you gave TikTok.
What changed in TikTok's terms in 2026?
The license changed more than the privacy policy did. On January 22, 2026, the date the US terms first named TikTok USDS Joint Venture LLC, the license began to mention machine learning, messages, and AI prompts. The privacy policy had listed machine learning as a purpose since at least 2024.
| Document and date | Training language | Messages | Who you contract with | Source |
|---|---|---|---|---|
| US terms, November 2023 | None in the license | Not in the license | TikTok Inc. | Archived December 1, 2025 |
| US privacy policy, August 19, 2024 | Training machine learning models listed as its own purpose | Scanned "for violations of our Terms of Service, Community Guidelines, or other conditions" | Not applicable | Archived December 1, 2025 |
| US terms, January 22, 2026 | "Including training, testing, and improving our machine learning models and algorithms" | Part of Your Content, "including messages" | TikTok USDS Joint Venture LLC | Archived February 1, 2026 |
| US privacy policy, January 22, 2026 | The same purpose, reworded to add testing | Scanning moves into the same sentence as training | Not applicable | Archived February 1, 2026 |
| US terms and privacy policy, July 15, 2026 | Unchanged from January | Unchanged from January | TikTok USDS Joint Venture LLC | US terms, US privacy policy |
The November 2023 terms already granted an "unconditional irrevocable, non-exclusive, royalty-free, fully transferable, perpetual worldwide licence" to your User Content. In January 2026, the license itself gained the training purpose, and messages and AI prompts moved inside it.
Location changed too. The 2024 policy said current versions of the app "do not collect precise or approximate GPS information from U.S. users." The July 2026 policy says TikTok "may also collect precise location data, depending on your settings."
Can you opt out of TikTok AI training?
Not in the US. Neither the US terms nor the US privacy policy names a training setting. The license is "subject to your Platform settings," and none of those settings mentions training.
In the EEA, the UK, and Switzerland, you can object. The EEA privacy policy, last updated November 30, 2025, names TikTok Technology Limited in Ireland and TikTok Information Technologies UK Limited as joint controllers, and lists "training, testing, and improving our technology, such as our machine learning models and algorithms" under its legitimate interests. TikTok's right-to-object page says to use the privacy webform and pick "Make an objection to the processing of my data." It asks you to name "the processing activity you are objecting to" and explain "how the processing activity affects you."
TikTok then runs a balancing test. The page says an objection can fail if you haven't "provided specific grounds," or if TikTok decides your interests "do not outweigh our legitimate interests." If it refuses, you can complain to the Irish Data Protection Commission or your local regulator. So name model training as the activity, and say why it affects you.
Does TikTok own what you make with its AI tools?
You own your Output, but only yours. The US terms define Output as the "responses those interfaces generate," and say "your ownership of Output does not extend to other users' output," because "output may not be unique to a specific user." Output falls inside Your Content, so the same training license covers it.
The terms also bar you from presenting Output as "human-generated," and from removing "any watermarks, content-authenticating metadata, or other marking or disclosure." TikTok's AI Alive post says moderation reviews "the uploaded photo and written AI generation prompt" before you see the video. Stories made with it carry an AI-generated label and C2PA metadata, which "helps others identify that it's AI-generated, even if it's downloaded and shared off platform." For how other AI apps split ownership of what you type and what comes back, see who owns your AI chats.
Does deleting your TikTok account remove your videos from its AI?
Not in the US. Section 5.1 of the US terms keeps the content license alive after you delete your account. The EEA terms end it when you close the account, unless you let others reuse your content or TikTok must keep it "for legal reasons." Neither version says a model already trained on your videos will be retrained without them. CapCut, run by the same US joint venture, takes a perpetual license that also outlives your account, which we cover in does CapCut use your videos for AI.
How Morse keeps your questions out of TikTok's training clause
TikTok's license reaches anything you hand to its AI-powered interfaces: the prompt for an AI Alive story, the question you'd type to make sense of a health trend in your feed, the caption you ask it to fix. Each becomes Your Content under an irrevocable license that includes training. To keep them out, ask somewhere that doesn't keep your words or train on them.
Morse is an AI chat built that way. Your chat history is stored only in your browser, encrypted, not on Morse's servers, and Morse doesn't train AI on your content, sell your data, or show ads (privacy policy). There's no training switch to hunt for and no objection form to argue. Each chat runs in the privacy mode you pick:
- Private, for the questions a video left you with. It's the mode every new chat opens in. Your message goes to Grok or an open-weight model such as DeepSeek, Kimi, or GLM, on hosts that agreed to keep nothing under Morse's zero data retention agreements. Where a host confirms zero retention on each reply, Morse checks and refuses a reply that doesn't.
- Anonymous, to ask GPT, Claude, or Gemini without them knowing it's you. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider may keep it for a limited time to check for abuse, under its own policy, so leave out names you'd rather not share.
- E2EE, on Pro, for anything no one should read. Your browser checks the enclave's hardware attestation, then encrypts the message so only that enclave can read it, and Morse's servers relay ciphertext they can't open. That rests on the enclave hardware and attestation working as designed.
Your history locks when you sign out and takes two keys to open: one that never leaves your device and one our server gives you at sign-in. Clearing your browser deletes it for good, and Morse can't recover it, so keep a copy of anything you need. Because we don't store chats, we usually have very little to give authorities.
What to do on TikTok today
- Training, in the US: there's no setting. Treat anything you post, draft, message, or prompt on TikTok as covered by the license.
- Training, in the EEA, UK, or Switzerland: object through the privacy webform, name model training, and give your reasons.
- DMs: keep anything sensitive out of them. They're inside Your Content, and they aren't end-to-end encrypted.
- Leaving: download your data first, under Settings and privacy, Account, then Download your data (TikTok). In the US, the license survives deletion. In the EEA and UK, it ends with your account, with the two exceptions above.
- AI help with a caption, a script, or a question: ask it in Morse, where your chat is kept in your browser and never used for training. For more on keeping chats off company servers, see how to use AI privately.
Ask your next question from the Morse home page without an account, or start free with no card. End-to-end encrypted chats come with Pro, at $19 a month.
Common questions
Does TikTok sell my data?
TikTok's US privacy policy, updated July 15, 2026, says it does not sell your personal information or share it for cross-context behavioral advertising "where restricted by applicable law." It still uses your activity for ads, and the **Ads** section of the app's settings lets you manage how your activity elsewhere online is used.
Are TikTok DMs end-to-end encrypted?
No. TikTok told the BBC in March 2026 that it would not add end-to-end encryption to direct messages, TechCrunch reported on March 4, 2026, and said DMs use "standard encryption, similar to services like Gmail," with access limited to authorized employees.
Does TikTok record what I type?
Its US privacy policy lists "keystroke patterns or rhythms" among the technical information it collects automatically, along with your device model, network type, and battery state. The same policy covers clipboard text and images when you choose to paste or share through the app.
Can I see less AI-generated content on TikTok?
Yes, in some accounts. TikTok's newsroom said in November 2025 that it would test a control in Manage Topics to set how much AI-generated content appears in your For You feed, which tailors the feed rather than removing such videos. It has no effect on whether TikTok trains on your own posts.