Canva is receiving backlash for racism—especially in regard to Black hairstyles—in its latest technological update that was meant to improve user experience.
In November 2022, the graphic design company announced its Text to Image app, an image-generating feature that utilizes text and artificial intelligence to curate an image exactly to the creator’s liking.
As for the safety precautions installed in its algorithm, Canva described how it “invested heavily” to limit the creation of unsafe content, saying, “For Text to Image this includes automated reviews of input prompts for terms that might generate unsafe imagery, and of output images for a range of categories including adult content, hate, and abuse.”
However, what Canva did not account for is racial biases appearing in its search results.
Adriele Parker, a DEI thought partner, detailed her experience using Canva while trying to find photos of Black women in popular hairstyles. She was caught by surprise
by what Canva’s search algorithm listed as “unsafe.”In a LinkedIn post, Parker shared the screenshot that displayed how an error occurred as she typed in the phrase “Black woman with Bantu knots.”
The digital creative platform also stated that “bantu may result in unsafe or offensive content.”
In her caption, Parker continued on how the likely lack of Black women in Canva’s technology department led to the prejudice.
“Tell me your AI team doesn’t have any Black women without telling me your AI team doesn’t have any Black women,” she vented.
“Canva, if you need a DEI consultant, give me a shout. I’ve been a fan of your platform for some time, but this is not it. Be the change. Please.”
Canva’s team responded as the post gained considerable attention and more users received the same result for similar race-based texts. Canva apologized for the “inconvenience” while emphasizing that its safety features required feedback such as Parker’s to ensure they were reasonable.
The response by the tech company was, in Parker’s opinion, a regurgitated and “canned” apology that did not address the issue at hand.