Canva’s AI Tool Decided ‘Palestine’ Needed Replacing — With ‘Ukraine’

I suddenly feel so much better about every embarrassing typo I’ve ever made. Because at least I’ve never had an AI tool unilaterally decide to swap a country’s name in my design.

Canva’s new Magic Layers feature — the one that’s supposed to break flat images into separate editable components — got caught doing something it definitely shouldn’t. X user @ros_ie9 discovered that when she uploaded a design containing the phrase “cats for Palestine,” the AI quietly changed it to “cats for Ukraine.”

Let that sink in. This isn’t a translation error. It’s not a misread character. The feature, which is meant to separate layers without altering content, specifically targeted the word “Palestine” and replaced it. @ros_ie9 noted that related terms like “Gaza” went untouched. Just “Palestine.”

Canva says it has now resolved the issue and is taking steps to prevent it from happening again. “We became aware of an issue…” — you know the drill. The standard corporate apology template.

But here’s what bothers me: this wasn’t a hallucination or a garbled output. This was a deliberate substitution. Somewhere in Canva’s training data, moderation filters, or prompt engineering, someone decided that “Palestine” should be replaced. That’s not an accident. That’s a policy decision hidden inside a feature that’s supposed to be invisible to users.

I’ve been watching these AI content moderation issues pile up for years. Google’s Gemini refused to generate white people. Midjourney had its own political filters. But this one feels different because Magic Layers isn’t a generative tool — it’s supposed to be a utility. You don’t expect your layer separation tool to rewrite your text.

Canva is probably going to blame it on a training data imbalance, or an overzealous safety filter, or some intern’s bad regex. And maybe that’s true. But the pattern is getting old. Every time an AI tool does something like this, we get an apology and a fix, but we never get a clear explanation of how the decision was made in the first place.

If you’re using Canva for anything politically or culturally sensitive, I’d suggest double-checking every output from Magic Layers for a while. Or just don’t use it until Canva explains exactly what happened and how they’re auditing for similar issues.

Because if an AI can silently swap “Palestine” for “Ukraine,” what else is it changing that nobody has noticed yet?