Canva's AI off switch: where it is, and what it leaves on

Share
Canva's AI off switch: where it is, and what it leaves on

Canva shipped a switch that turns its generative AI off. It is the first time a mainstream design platform has admitted, in product form, that a large share of its users would rather not have the feature at all. The revealing part is not the toggle. It is the line Canva drew around what still runs with it off.

Canva's generative AI switch is live for every free and paid user as of Thursday, under Profile > Settings > Permissions, and turning it off hides Magic Write, Magic Design, Magic Layers, Magic Edit, Magic Expand, Dream Lab, Canva Code and AI Shapes across the product, according to Canva's announcement. The Verge's Emma Roth, who reported the change first, notes that the switch also blocks image generation through Canva's chatbot. Existing designs are untouched: the toggle governs what you can do next, not what you already made, and it can be flipped back on at any time. There is no upsell attached — the controls ship to Canva Free and Canva Pro at no extra cost.

The exception list is where the editorial thinking shows. Background Remover and Magic Grab stay available with generative AI switched off. Canva's reasoning, in its own words, is that both tools "only work with what's already in a design" — they edit existing pixels rather than inventing new ones.

That sounds like a technical distinction. It is really a labeling decision. Canva says that because those tools do not generate anything new, using them "won't trigger an AI-generated content label," and the company adds the consequence explicitly: social platforms that read Canva's file metadata "won't label your work as AI-generated simply because you've used tools like Background Remover."

Read that twice, because it is the most consequential sentence in the announcement. Background removal is machine learning. Magic Grab, which lets you lift an object out of a photo and reposition it, is machine learning. What separates them from Magic Expand is not whether a model ran — it is whether the output counts as synthetic content under the labeling rules that Instagram, YouTube and others are now enforcing. Canva has not built an AI-free mode. It has built a labeling settlement, and it is telling designers which side of it each tool sits on.

The second control is a discovery switch rather than an editing one. Under Profile > Your account > AI Personalization, users can ask Canva to show less AI-generated content while browsing the template and element library. Canva says the option reduces AI-made assets in search results and recommendations and brings human-created work forward, without changing designs you have already built. AI-generated assets in the library already carry a metadata label, part of what Canva calls broader industry efforts on transparency, and creators are paid when their work is used — a detail worth remembering, because the switch is partly a creator-relations move as much as a user-preference one.

Why now is not mysterious. Canva says its community has used the company's AI tools more than 38 billion times, and it has spent three years pushing in one direction: the Magic Studio launch in 2023, prompt-powered editing in April, and Canva Craft, its first in-house image model, last month. Along the way, users on Canva's own forums and subreddits asked, repeatedly and sometimes angrily, for a way to disable the features. Roth links several of those threads.

The business logic of shipping the off switch anyway is the interesting part. Canva's bet for years has been that AI is the product — the reason a free user upgrades and a team standardizes. An off switch looks like a concession to that bet. In practice it is retention triage: the user who quits over feature bloat is worth less than the user who stays with AI hidden, and the default stays on. Nothing in the announcement suggests Canva expects many people to flip it. What it buys is the ability to say nobody is forced.

There are limits worth naming, because the framing invites overreading. AI remains on by default; you have to go three menus deep to find the switch, which is the standard dark-pattern-adjacent place to put an opt-out. Designs that already contain AI-generated assets keep their labels — the toggle does not launder a cover made with Dream Lab. Organizations on Canva Business, Enterprise, Teams and Education already had admin-level AI permissions; the new controls extend that logic down to individuals rather than inventing it. And the announcement says nothing about training, retention or how your content is used, so treat these as display and access controls, not privacy controls. If your question is what Canva does with your uploads, this post does not answer it.

The honest counter-argument is that a settings-page toggle is a weak substitute for not shipping the bloat in the first place. Users who asked Canva to stop adding AI panels did not ask for a menu item to hide them. That is fair, and it is also how every mature software suite handles a contested feature: the loud minority gets a switch, the default keeps the revenue. What matters is whether the switch is real, and this one appears to be — the features are hidden, not merely collapsed, and the metadata behavior is documented.

Seen from outside Canva, the more useful signal is the direction of travel. For two years, AI features arrived as weather: unavoidable, on by default, impossible to decline. Now the off switch is becoming a selling point in its own right. Google made AI off by default in its classroom products and then found Search could not be separated from it, a tradeoff we covered in Google schools: AI off by default — and Search can't be had without it. Canva has taken the easier version of the same step: it sells the switch without surrendering the default. Enterprise admin panels have had this for a while. What is new is consumer software treating "you can turn it off" as a feature worth announcing.

If the toggle measurably cuts AI usage and churn does not move, expect the pattern to spread: suites will compete on control rather than capability, and "made without AI" will become a filter, a badge, maybe a premium tier. If it cuts usage and churn is unaffected either way, the controls quietly stay buried in settings and nobody announces the next one.

What to watch: whether Canva publishes any usage data for the switch, whether Adobe and Figma match it with end-user controls rather than admin ones, and whether the metadata label becomes the thing clients actually check before accepting a deliverable.

Would you pay more for a design tool that promises no AI in the loop, or is a hidden toggle enough? Write to us via our contact page.

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.

Read more

Apple Health AI study: 200,000 iPhone owners, one year, no diagnosis

Apple Health AI study: 200,000 iPhone owners, one year, no diagnosis

Apple did not ship a health chatbot this week. It shipped a study — and the study is the most concrete signal yet of how the company intends to enter consumer health AI. Apple opened enrollment Monday for the Apple Health AI Usage Study, a research app that lets U.S. iPhone owners ask everyday medical questions and rate the answers, targeting up to 200,000 participants over more than twelve months. The app is already live on the App Store, free, and lists Apple as the developer.

Apple Intelligence in China: Qwen comes to Siri on November 4

Apple Intelligence in China: Qwen comes to Siri on November 4

Two years after the rest of the world got it, Apple's AI has a launch date in the company's biggest market outside the United States — with Alibaba's model riding along, and the rebuilt Siri left at the border. Apple will bring the first batch of Apple Intelligence features to China on November 4 as a developer beta, with the public release in December alongside iOS 27.3, and Alibaba's Qwen will be available inside Siri, the writing tools and Visual Intelligence.

Vibe-coded game ports put Halo and GTA in the browser

Vibe-coded game ports put Halo and GTA in the browser

Classic console games are collapsing into the web at a pace nobody asked for: in the past two weeks, hundreds of AI-decompiled ports of PS2- and PS3-era titles have appeared online, and several of them play like the real thing. Hundreds of AI-decompiled games — Halo: CE, GTA: Vice City, Call of Duty: Black Ops, Skate 3, The Simpsons: Hit and Run — now run in a browser tab, and the ones we've seen reports of don't look like bootleg garbage.

Goncourt drops the prize novel accused of being AI-written

Goncourt drops the prize novel accused of being AI-written

France's literary season just produced its first casualty of AI, and nobody involved can actually prove authorship either way. The Académie Goncourt withdrew Thélyson Orélien's novel C'était ça ou mourir from its selection on Friday 25 September, four days after the book won the Prix du Roman Fnac. The academy said the decision was driven by "the plagiarism of which its author has been guilty over time" and by "the convergent result of several analyses by credible researchers and journalists"…

Meta's hologram avatars put a photoreal you on video calls

Meta's hologram avatars put a photoreal you on video calls

Meta spent Connect this week selling hardware. The announcement that will outlast the hardware is the face it puts on your calls: at Wednesday's keynote the company said its holographic avatars — the productized version of the Codec Avatars research it has demoed since the Orion prototype — are coming to the Meta VR Glasses and to the Ray-Ban Display frames. The avatars are photoreal, they run on a model Meta calls Muse Realtime Avatar, and the first place you will meet one is a WhatsApp call.

XPRIZE's wildfire AI found fires in 10 minutes — couldn't stop them

XPRIZE's wildfire AI found fires in 10 minutes — couldn't stop them

The $11 million XPRIZE Wildfire handed out its awards on Wednesday, and the honest headline is the one the organizers didn't write: the detection half worked, the firefighting half didn't. Meanwhile Black Forest Labs quietly put a robot model at the top of Nvidia's leaderboard. XPRIZE awarded $4.55 million in prizes at a ceremony in New York on Wednesday, and both $3.5 million grand prizes went unclaimed because no team could finish the job it was actually set. The competition ran two tracks.