Canva's Magic Layers AI Silently Swapped 'Palestine' for 'Ukraine' in User Designs

Canva’s Magic Layers AI feature was silently rewriting user content — replacing the word “Palestine” with “Ukraine” in designs — without any user input or visible disclosure. The tool, whose stated purpose is to decompose flat images into separately editable layers, has no design mandate to alter text. That it did so, and did so for a single politically charged word, is a significant credibility problem for a company that just declared its AI overhaul “the beginning of the next era of creation.”

A Suspiciously Specific Substitution

What makes this incident more than a routine software glitch is its narrow scope. According to The Verge, X user @ros_ie9 — who surfaced the issue publicly — found that adjacent terms such as “Gaza” passed through the feature untouched. The substitution was precise: one word, one replacement. That pattern doesn’t fit a generic OCR or parsing error; it fits a content filter or a training artifact that had been inadvertently embedded into a production feature.

Canva spokesperson Louisa Green told The Verge the company “moved quickly to investigate and fix” the problem and is implementing “additional checks” going forward. The statement is diplomatically worded but sidesteps the structural question: how did a word-level substitution rule end up inside a layer-decomposition tool in the first place?

The Hidden Cost of AI Feature Bundling

The broader pattern here deserves scrutiny. As AI capabilities get bundled into design and productivity platforms — Canva’s Magic Layers, Adobe’s Generative Fill, Google Docs’ Smart Compose — content policies that were once visible settings become invisible infrastructure. Users don’t sign up for editorial curation when they use a layer-separation tool; they expect fidelity, not substitution.

Canva is directly competing with Adobe’s increasingly AI-heavy creative suite, and Magic Layers is a flagship differentiator in that battle. A bug that rewrites politically sensitive text erodes exactly the trust that enterprise and professional users require.

Why This Matters

This incident is a preview of a tension that will only grow louder: AI-assisted creative tools embed policy decisions at the model or pipeline level, making them functionally invisible to end users. When those decisions touch geopolitical language, the opacity becomes untenable. The question Canva hasn’t yet answered — what mechanism caused the substitution — is the one the industry should be pressing. Without that transparency, “we’ve fixed it” is a patch, not a safeguard.