Does Pika train on your data?
Pika's terms say it may use "anonymized or aggregated" Inputs, Outputs and interactions to train its models; personal data in API inputs "will generally be used" to train. No way to decline.
No opt-out found in this policy
"Unless otherwise agreed in writing, you acknowledge that Pika may use anonymized or aggregated Content, including Inputs, Outputs, and your interactions with the Service, and Usage Data to train, improve, and develop its machine learning models, algorithms, and related technologies and services"
- 21 Sept 2026
- 5 Oct 2026
- All accounts
- global
- Fetched, browser-rendered
- 0047DA48…EA07
How we read it: Re-read on 22 September 2026 after Pika rewrote both documents on 21 September, five days after this site's letter of 16 September, which drew no reply. The terms' training clause gained "anonymized or aggregated" before Content, and the privacy policy gained a sentence the other way: personal information in Inputs sent through the API platform "will generally be used to improve and train artificial intelligence models". De-identification is recorded as qualified for that reason. Unchanged: the likeness, voice samples and personal characteristics behind an AI Self train only that AI Self and are deleted with the account; no route to decline; enterprise agreements govern their own terms.
opt out, opt-out, object, consent, agreed in writing, delete, train, default
What the document does offer
- in the terms, the sentence after the clause: "Notwithstanding the foregoing, any personally identifiable data included in the Inputs used to create an AI Self (including your likeness, voice samples, and personal characteristics) will only be used to develop, operate, improve and train your AI Self and will not be used to train other Models or create AI representations for other users."
- in the terms: "Upon deletion of your Account, Pika will permanently delete all data" associated with the AI Self
- in the terms: "Opt-Out of Arbitration", and, since the 21 September rewrite, an opt-out from being named as a customer in Pika's marketing: "You may opt out of this use by notifying Pika in writing at support@pika.art". Neither concerns training
- in the privacy policy (pika.art/privacy-policy, captured 2026-09-21): "Where permitted by applicable law or with your consent, we may also use content you submit-including Inputs and Outputs-to improve our artificial intelligence models and related technologies."
- in the privacy policy: opt-outs for behavioural tracking, promotional email and targeted advertising, and the state-law "Request to opt-out of personal data processing for" list, which does not name training
- in the privacy policy, added 21 September 2026: "personal information that you include in Inputs through the API platform will generally be used to improve and train artificial intelligence models"
- "default": 0 matches in either document
Why none of it applies: No route is stated. "Opt out" appears in the terms for arbitration and for being named in Pika's marketing, and in the privacy policy for behavioural tracking, promotional email, targeted advertising and state-law rights, none of which names training. The terms' condition, "unless otherwise agreed in writing", is a contract to negotiate, not a control to use; the privacy policy's "Where permitted by applicable law or with your consent" describes the law. Deleting the account deletes the AI Self's data: an exit, not an opt-out. Scored as no opt-out. The 21 September rewrite added no route.
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.
- None offered
- Not stated
- 88 / 100
- Qualified
- Unknown
How to opt out of Pika AI training
Plan by plan, from what Pika'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
No opt-out is stated in the policy on file for Pika.
What this policy has done since we started watching
2 changes detected since 6 Sept 2026. Each one is a diff between two archived copies, and each links to the passage that moved.
Vendor response
We wrote to Pika on 16 Sept 2026, quoting the clauses above and inviting a correction. Nothing has come back. That's published because a company declining to comment on its own policy is something a reader should know, and because the invitation should be on the record whether or not it's taken up. Any reply goes here, word for word.
Report a wrong clause