Does Figma train AI on your designs? The admin switch, by plan

The Morse teamPublished October 29, 202610 min read

Yes, as of October 29, 2026, Figma does train its AI on your designs when they sit in a Starter or Professional team whose admin hasn't turned off Content training, and it doesn't for Organization, Enterprise, Education, or Government accounts (Figma). Only an admin can change the setting. Turning it off covers new content and edits from that day on, and it doesn't reach free files you publish to the Figma Community, which Figma's own models were trained on in part.

We read Figma's help article on AI settings (updated October 27, 2026), its Building Figma AI page, its AI Terms, privacy policy, Terms of Service, list of sub-processors, AI FAQ, and AI credits page, its 2024 blog posts about Figma AI, and the court's May 2026 order in Khan v. Figma, all on October 29, 2026. Figma deserves credit in two places. It states the training default for every plan in plain words, which Canva, for one, doesn't do for personal accounts. And it says it doesn't let OpenAI or its other model providers train their own models on what you make in Figma.

Which Figma plans are used for AI training?

Starter and Professional teams, unless an admin opts out. Each row comes from Figma's own pages, read on October 29, 2026.

Figma explains the split between plans on its approach page: "Our customer agreements with Organization and Enterprise are typically more complex and include specific requirements and restrictions, which is why we've chosen a different default setting for those plans." On Starter and Professional the switch is set per team, and on Organization and Enterprise it's set for the whole organization.

How do you turn off content training in Figma?

Ask an admin, or be one. Figma's help article gives two paths:

• Starter: In the file browser, open your team and click All Folders in the left sidebar. Open the menu next to your team name, choose View settings, go to the Settings tab, scroll to the AI section, select Manage AI settings, and toggle Content training off.
• Professional: In the file browser, open your team and click Admin in the left sidebar. Open the Settings tab, scroll to AI, and toggle Content training off.

A member who isn't an admin has no switch of their own. For drafts, Figma's page offers two ways out: turn off content training "or ask your admin to do so," or move your drafts into a team where it's off. It says "drafts organized into teams are treated the same as other files in that team," so the setting follows the team a file lives in. A freelancer's work inside a client's Professional team follows the client's switch.

What does Figma's content training include?

Anything made in or uploaded to Figma. Figma's approach page lists text and images, comments and annotations, and layer names and properties as examples of customer content. What Figma AI generates for you counts as well. The same page says AI output "will not be used to train models unless the content training setting is 'on'."

Usage data is handled apart from the switch. Figma describes it as information about how your content is used ("like how many times it is accessed"), technical logs and metadata, and telemetry, "used in an aggregated, de-identified way." Figma's pages name no setting for it.

Before training, Figma says it takes steps so its models "learn general design patterns and Figma-specific concepts and tools" rather than "your content, concepts, and ideas," and gives an example: "we de-identify content and redact sensitive information including from text and images" (Figma). Its AI Terms, effective June 24, 2026, put the permission in one line: when the toggle is on, "Figma may use Customer Content to maintain, improve, and enhance Figma's products and services by training machine learning and artificial intelligence algorithms and models." Figma's privacy policy, effective September 1, 2026, gives its legal basis for this as its legitimate interests, "to keep our Services updated and relevant," rather than your consent.

Does turning off content training remove what Figma already used?

No, going by Figma's own wording. The AI Terms say that once an admin turns the toggle off, "new Customer Content and edits added after the date the toggle was turned off will not be used to train AI models." None of the pages we read says content already used is taken back out of a model.

Content training "went into effect on August 15, 2024" (Figma). When Figma announced it on June 26, 2024, it promised "No content training will occur until August 15, so you have time to decide" (Figma). June 26, 2024 is a cutoff for one group only, the teams that had turned off the beta AI setting before that day, which started with content training off. It was never an opt-out deadline, since Starter and Professional admins can turn the setting off "at any time." A Starter or Professional team whose admin never opened the setting has been open to training since August 15, 2024.

Figma's AI Terms page lists seven versions since June 26, 2024, the latest effective June 24, 2026 (Figma).

Does Figma train AI on Community files?

Yes, on the free ones, and the content training switch doesn't cover them. Figma's AI FAQ, updated October 23, 2026, says "The first-party models used across Figma were trained in part on data sourced from publicly available datasets published on the Figma Community website and licensed under CC BY 4.0." The approach page adds that "generative AI features are fine tuned with data from public, free Community files," with Visual Search, Asset Search, and Add Interactions named, and that "we do not use paid files to train AI models."

That reverses what Figma said in June 2024: "Until we decide and transparently communicate our approach, generative models that output design will not be trained on Community files" (Figma). Some opt-out guides still repeat the 2024 line, including Redact's, dated September 2026. Free Community files carry a CC BY 4.0 license, which lets anyone adapt them "for any purpose" with credit (Figma). Figma says it credits the Community on its Help Center. We found no setting on Figma's pages that keeps a published free file out.

Do OpenAI and other AI companies train on your Figma files?

Not by Figma's account. Figma says it does "not permit 3rd party model providers to use data that customers upload or create on the Figma platform for training their own models," and that "OpenAI and our other 3rd party model providers store data temporarily, or in some cases not at all" (Figma). It gives no period, and sends you to each provider's own policies for one.

Figma's list of sub-processors, updated October 23, 2026, names ten AI model providers for Figma AI: Anthropic, Amazon Web Services, Cerebras, Fireworks.ai, Google (Vertex), Jasper, Microsoft (Azure), OpenAI, QuiverAI, and Recraft. Nine are listed in the United States with a footnote: requests may be routed through global data centers, where data is processed "exclusively in volatile memory" and the cache "is automatically cleared within two hours." QuiverAI is listed in the United States and the EU. Weave tools inside Figma draw on a separate list of 15 providers, among them Kling AI, BytePlus, Runway, and ElevenLabs. Figma's AI credits page shows you choosing between models by name for image tools, ChatGPT Images 2.0 and the Nano Banana models among them.

What happened with Figma AI in 2024 and since?

Figma's AI launch, its pulled feature, and the lawsuit that followed, by date:

• June 26, 2024: Figma announces Figma AI at Config, including Make Designs, and the content training setting. It says its generative features "were not trained on private Figma files or customer data" (Figma).
• July 2, 2024: After designer Andy Allen shows Make Designs returning designs very close to Apple's Weather app, CEO Dylan Field disables the feature. He writes on X that "the accusations around data training in this tweet are false" and "Ultimately it is my fault for not insisting on a better QA process for this work," as TechCrunch reported (TechCrunch).
• July 17, 2024: Figma's retrospective says Make Designs used off-the-shelf models, "OpenAI's GPT-4o and Amazon's Titan model," and "we have not done any additional training or fine-tuning." It traces the resemblance to design system components added the week before Config "that we simply didn't vet carefully enough" (Figma).
• August 15, 2024: Content training takes effect (Figma).
• September 24, 2024: The feature returns in limited beta as First Draft. Figma says "First Draft doesn't train on customer content" (Figma).
• July 24, 2025: Figma AI becomes generally available and the beta on-off setting is removed. Teams that had switched AI off stay off until an admin turns it back on (Figma).
• November 21, 2025: Raza Khan files a proposed class action alleging Figma trained AI on user designs without consent, as Bloomberg Law reported (Bloomberg Law).
• May 8, 2026: The court sends the case to arbitration and stays it (order).

Is Figma being sued over AI training?

Yes, though the case is on hold. In Khan v. Figma, the complaint says Figma "never asked its customers for permission," according to Bloomberg Law. The plaintiff's law firm says Figma changed the "default setting" for millions of accounts to opt them in without fair notice (Greenbaum Olbrantz). These are allegations, and no court has ruled on them.

On May 8, 2026, Judge Charles R. Breyer of the Northern District of California granted Figma's motion to compel arbitration and stayed the case (order). The plaintiff argued his claims fall under the terms' exception for protecting intellectual property in court. The judge left that question to the arbitrator, noting "Khan's strong arguments regarding the carve-out." Figma's current Terms of Service, effective March 11, 2026, still send disputes to binding arbitration, except that either side may sue in court "to enjoin the infringement or other misuse of intellectual property rights." This isn't legal advice.

How Morse keeps your design questions out of training

In Figma, whether your work trains a model comes down to a switch someone else holds: your team's admin, or your client's. Anything you publish free to the Community is used regardless of the switch, and turning it off doesn't take back what was used before. Morse has no such switch, because Morse doesn't train AI on your content, sell your data, or show ads (privacy policy).

Keep your design files in Figma. The questions you'd rather keep out of anyone's training set, such as the brief for an unannounced feature, the critique you're drafting for a client, or a product name you're testing, can go to a Morse chat instead. Your chat history is stored in your browser, encrypted, and so are the images you make in Studio, Morse's image tool. Each chat runs in the privacy mode you pick:

• Private, for briefs, UX copy, and critique notes. New chats start here. Grok and open-weight models such as DeepSeek, Kimi, and GLM run on providers that agreed to keep nothing, since Morse has zero data retention agreements with its inference, search, and other providers. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, on Pro, for a launch or a client no one should hear about yet. Your browser checks the enclave's proof that it's genuine before anything leaves your device, then encrypts the message so only that enclave can read it. Our servers relay text they can't read. That holds as long as the enclave hardware and attestation work as designed. E2EE has no image generation or web search, so settle the words there first.
• Anonymous, for asking GPT, Claude, or Gemini without the provider 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 sees the message and may keep it for a limited time to check for abuse, under its own policy, so leave client and product names out.

History in every mode is locked when you sign out and opened with two keys, one of which never leaves your device. Clearing your browser deletes your chats and Studio images, and Morse can't recover them, so keep anything you'll need. Because we don't store chats, we usually have very little to give authorities.

What to check in Figma today, by your role

• Starter or Professional admin: Turn off Content training in your team's AI settings. It stops new content and edits from training from that day.
• Member of someone else's team: Ask the admin which way the switch is set, or move your drafts into a team where it's off.
• Freelancer working in a client's team: The client's setting applies to your files there, so ask before you upload anything under NDA.
• Community creator: Free files you publish are training data for Figma's models, and Figma names no switch for them.
• Organization or Enterprise: Training is off and locked; Figma says "we'll let you know if this changes."
• The words around the design: Draft briefs and critiques in a Morse chat, since Morse doesn't train AI on anything you write. For other apps' rules, see which AI companies train on your data and how Adobe's terms treat your files.

Start a chat from the Morse home page with no account, or sign up free, with no card, for 15 messages and 3 Studio images a day. Pro, at $19 a month, adds E2EE chats.