Does Recraft train on your data?
Recraft trains its own models on the images and prompts you submit. Recraft says the opt-out toggle works on Pro, is applied for Team plans, and is not offered on the Free plan at all.
Opt-out is sold, not offered
Graded on the plan most people are on. The other plans are in the matrix below.
"We may use such information to train Recraft's own proprietary AI models, but you may disallow such use of your information by opting out in your account profile settings."
- 14 Aug 2026
- 5 Oct 2026
- Free
- global
- Fetched from the company
- 964F1601…A84F
How we read it: Recraft's support reply of 25 Aug 2026 states this plainly: "Training opt-out is not available on the Free plan." Nothing in the privacy policy says so. The policy's one sentence on the subject offers the opt-out without qualification, which is why this page previously described the switch as free and named — that was our inference from an unqualified sentence, and the company says it is wrong. Graded E rather than D because the tier has no way out at all: the control exists, is described in the same policy that governs this plan, and is reachable only by paying. That is the privacy paywall in its plainest form.
free, plan, Pro, subscription, opt out, profile settings
What the document does offer
- the policy's opt-out sentence names no plan and carries no qualification: "you may disallow such use of your information by opting out in your account profile settings"
- "Free" appears in the privacy capture only in "free of charge", about the licence granted to Recraft, not about plan tiers
- no plan name — Free, Pro or Team — appears anywhere in the privacy policy
Why none of it applies: The published policy describes one opt-out available to everyone. Recraft's support team says that is not the case, and that the Free plan cannot reach it. Both cannot be true. This tier is graded on the company's own statement, which is the more restrictive of the two, and the discrepancy is recorded rather than resolved: the sentence a Free user would actually read still tells them a control they do not have is theirs.
"We may use such information to train Recraft's own proprietary AI models, but you may disallow such use of your information by opting out in your account profile settings."
- 14 Aug 2026
- 5 Oct 2026
- Pro
- global
- Fetched from the company
- 964F1601…A84F
How we read it: A clean, self-contained sentence: it states the training use and the control in the same breath, and names where the control lives rather than deferring to a help article. Recraft's reply of 25 Aug 2026 adds the switch's label — "Help improve Recraft", in Profile settings — and says opting out applies to future inputs, not to what has already been used. Graded D rather than C because the policy does not surface the setting at signup and does not say whether it starts on or off.
opt out, opting out, profile settings, train, withdraw, by default, default
What the document does offer
- "We may use such information to train Recraft's own proprietary AI models, but you may disallow such use of your information by opting out in your account profile settings" — the control is in the same sentence as the training use, and names where it lives
- "Where we ask for your consent to process your information, you can always withdraw such consent"
- device, browser and geolocation settings, which govern collection
- "default" appears zero times in the capture
Why none of it applies: One sentence states the training use, the control, and its location, which is more than most documents here manage. The omission is the same one: nothing says whether the account profile setting starts on or off, and "default" is absent from the document. Recraft's reply supplies the answer the document withholds — on by default, opt-out forward-looking only — and that answer is published nowhere.
What Recraft said when we showed them this
Clarified“Recraft uses user-generated content—such as images created or uploaded, and text entered in prompts or chat mode—to improve its AI models by default. However, there is an opt-out control available: On the Pro plan, users can opt out of model training by turning off the "Help improve Recraft" toggle in their Profile settings. [...] Training opt-out is not available on the Free plan. For Team plans, opt-out is applied by default. API inputs and outputs are never used for model training.”
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.
- Paid upgrade required
- Not stated
- 76 / 100
- Not stated
- Unknown
- Self-serve toggle
- Not stated
- 53 / 100
- Not stated
- Unknown
- Not established
- Not stated
- —
- —
- Not established
- Not established
- Not stated
- —
- —
- Not established
How to opt out of Recraft AI training
Plan by plan, from what Recraft'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.
Free
Only by moving to a higher-priced plan. Recraft puts the opt-out behind a paid tier.
Pro
Yes. Recraft has a self-serve setting.
- Log in to Recraft on desktop.
- Go to your Profile settings.
- Near "Help improve Recraft", turn the toggle off.
Recraft's docs say the opt-out is not offered on the Free plan and that Team plans are opted out by default.
Team
Not graded, so no route is pointed at. The privacy policy names no plan tiers at all — not Free, not Pro, not Team — and says only that training may happen and that a profile setting can disallow it. So nothing published says whether a Team account starts opted in or opted out, and the one statement that does say is an automated support email. Grading B on that would convert a helpdesk reply into a published commitment, which is the conversion this site exists to refuse. Recorded as unclear until Recraft states it somewhere a Team customer could find it.
API
Not graded, so no route is pointed at. The privacy policy draws no distinction between data arriving through the API and data arriving through the app; its training sentence covers "such information" without qualification. A carve-out this broad would be worth a sentence in the document and does not have one. Left unclear rather than graded, because the alternative is to give Recraft credit for a commitment no API customer could cite back to them.
What this policy has done since we started watching
No change detected. We've held a copy of Recraft's documents since 14 Aug 2026 and re-read them since, across 3 captures of 2 documents. The text hasn't moved in a way that touches this verdict. That's a fact about the documents, and it's only knowable by having kept the earlier copies.
Vendor response
Their reply is quoted in full at the top of this page.
We wrote back on 25 Aug 2026 with a question that's still open. It's listed on the register.