Canva apologizes after its AI swapped "Palestine" for "Ukraine" in user designs

Canva has landed in the middle of an unexpected controversy: Magic Layers, one of the platform's new AI tools, was replacing the word "Palestine" with "Ukraine" in user designs on its own, without the creators' knowledge or consent. The feature was only supposed to split images into layers. Instead, it quietly edited politically sensitive text. The company acknowledged the problem, issued a public apology, and said the bug has been fixed, but it has not offered a detailed technical explanation of what happened.

Magic Layers is one of the flagship features of Canva's sweeping AI update, unveiled in 2026. It takes a flat raster image and breaks it into separate editable layers — background, main objects, text blocks — so designers can work with each element independently, without manual selection or retouching. By design, the tool should only restructure an image, not change its content: not the text, not the colors, not the placement of elements.

That made the discovery by X user @ros_ie9 all the more surprising. After running Magic Layers on an image with the phrase "cats for Palestine" — a common solidarity slogan — the user got back a design reading "cats for Ukraine." The swap happened automatically, with no warning, no confirmation dialog, and no explanation in the interface. The post demonstrating the bug spread quickly across social media and sparked wide discussion among designers and human rights activists.

The most telling detail was that the bug was selective. The user deliberately tested related terms — "Gaza," "Palestinian," "Palestine solidarity," and other geographic and political labels — and all of them were processed correctly and left untouched. Only one word was replaced: "Palestine" in its exact form. That immediately raised pointed questions: was this a random technical glitch in the language model, or had a content filter that specifically excluded this one word made it into production?

Canva's official response was brief and evasive. The company confirmed the issue and published a statement that began, "We became aware of an error that occurred..." — the full text was not widely shared. Canva assured users it had fixed the problem and was taking additional steps to prevent similar incidents. But no specifics followed: not which part of the pipeline failed, not how many designs may have been affected, and not why the replacement was so selective. That opacity only deepened public skepticism.

The potential scale of the problem is hard to overstate. Canva is one of the largest design platforms in the world, used by more than 220 million people, including human rights organizations, journalists, NGOs, and activists. When a tool with that kind of reach quietly edits a user's politically sensitive text — even as the result of a technical error rather than intent — it stops being just a bug. It becomes a question of trust, editorial independence, and users' right to control their own content.

There are several possible explanations. The model's training dataset may have contained samples in which "Palestine" was systematically replaced or flagged as undesirable, and the model learned that pattern. Another hypothesis is a bug in the OCR algorithm or the text normalization module that fails specifically on that particular spelling. A third possibility is that a content filter originally built for a different context accidentally made it into production. Without a transparent answer from Canva's engineering team pinpointing the exact source of the error, none of these theories can be confirmed or ruled out.

For designers who create content on political and humanitarian topics, the incident is a practical warning: AI features that look technically neutral and aren't meant to edit text can still change it without anyone noticing. After using any AI-powered tool, it's worth checking the final output especially carefully.