Canva has issued an apology after one of its newest AI features was caught quietly replacing the word “Palestine” with “Ukraine” in user designs, according to The Verge AI. The bug surfaced in Magic Layers, a flagship tool from Canva’s recent AI overhaul, and was first flagged by X user @ros_ie9, who showed the phrase “cats for Palestine” being silently rewritten as “cats for Ukraine.”
The Verge AI reports that the substitution appeared to target the word “Palestine” specifically. Related terms like “Gaza” passed through Magic Layers untouched, which makes the behavior look less like a generic content filter and more like a targeted swap baked somewhere into the model or its guardrails.
What Magic Layers Was Supposed to Do
Magic Layers isn’t a text editor. It’s designed to break flat images into separate editable components, so users can move, restyle, or recompose elements without rebuilding a design from scratch. Visible content changes weren’t part of the spec.
That’s what makes this incident more than a cosmetic glitch:
- The tool altered a politically charged word without warning the user.
- The change was silent, not flagged as an AI suggestion.
- It happened inside a feature Canva is positioning as central to its next-generation creative suite.
Canva’s Response
Canva spokesperson Louisa Green told The Verge the company moved fast once it caught wind of the reports.
We became aware of an issue with our Magic Layers feature and moved quickly to investigate and fix it. We take reports like this very seriously, and we’re putting additional checks in place to help prevent this in future. We’re sorry for any distress this may have caused.
Replies under the viral X post indicate other users reproduced the bug before the fix landed. The Verge AI’s own tests after the patch didn’t trigger any word swaps.
Why This Matters
Canva is in the middle of a serious push against Adobe’s AI-powered design stack, and Magic Layers sits at the center of what the company calls “the beginning of the next era of creation.” A trust-breaking bug in that exact feature is bad timing.
What stands out here is the broader pattern. AI design platforms keep shipping models with hidden behaviors that surface only when users stumble into them. Word substitutions, image filters, prompt rewrites, all of it tends to live deep in the stack, invisible to the person clicking the button. The Palestine-to-Ukraine swap is a reminder that AI tools aren’t neutral pipes. They carry choices, sometimes opaque ones, that can rewrite a designer’s intent without consent.
For practitioners, the takeaway is simple:
- Audit AI outputs against your originals, especially text inside images.
- Don’t assume “non-generative” features (layering, masking, upscaling) leave content untouched.
- Treat any AI-touched asset as something that needs a human review pass before it ships.
Canva says new checks are going in. Whether those checks catch the next variant of this problem, or just this one, is the real question. Full details at the original source.