Canva just had one of those AI moments that makes you wonder who's testing these things. The design platform's Magic Layers feature, which is supposed to break flat images into editable components without changing anything, was caught swapping the word "Palestine" for "Ukraine" in user designs.
A user on X noticed the issue when their "cats for Palestine" design mysteriously became "cats for Ukraine" after running it through the AI tool. The problem appeared specific to that one word, other related terms like "Gaza" worked fine.
This is the kind of AI behavior that raises immediate red flags. Either the training data had some serious bias baked in, or there was content filtering gone wrong. Neither explanation is great when you're running a platform used by millions of creators.
Canva responded quickly, saying they've fixed the issue and are working to prevent similar problems. But it's a reminder that AI tools can encode unexpected biases, even in seemingly straightforward features like layer separation.
For anyone using AI in their workflow, this is your periodic reminder to actually check the output. These tools are powerful, but they're not neutral. They make decisions based on their training, and sometimes those decisions reflect problems in how they were built.
