Generative AI features are being rolled out rapidly in a vast array of popular creative software. Perhaps sometimes a little too rapidly. It certainly seems that Figma could have performed a few more checks before launching its Make Design AI tool.
The UI design platform, which we rate as the best UI design tool, has disabled the new tool after it was found to apparently replicate the UI from existing apps, including the Apple weather app for iOS. It hopes to re-enable the feature soon, but this latest controversy involving a major software developer highlights the risk of rushing out AI tools, and it may put creatives off using them for fear of being accused of copying others' work.
Figma's Make Design is an AI-driven tool that's intended to allow users to create quick UI mockups. Not Boring Software CEO Andy Allen posted screenshots on X, showing that the tool would output a UI design for a weather app that was almost identical to Apple's existing app for iOS.
“Just a heads up to any designers using the new Make Designs feature that you may want to thoroughly check existing apps or modify the results heavily so that you don’t unknowingly land yourself in legal trouble,” Allen warned. It could be argued that weather apps are much of a muchness, but the side-by-side comparison throws up some uncanny similarities.
Figma CEO Dylan Field quickly announced on X that the tool had temporarily been disabled and blamed himself for pushing his team to get it out there fast. However, Figma has been unable to entirely dismiss the claim that Make Design was copying existing UIs because it doesn't know how the tool was trained.
Figma CTO Kris Rasmussen told Jay Peters of The Verge on Tuesday that the company didn't know how the generative AI was trained because that was done by a provider, who delivered "off-the-shelf models and a bespoke design system". On X, Rasmussen wrote: "We are looking into what extent the similarities are a function of the third party models we are using vs. the design systems we commissioned to be used by the models and we will address as needed."
It appears that those off-the-shelf models are OpenAI’s GPT-4o and Amazon’s Titan Image Generator G1, which raises questions about their training and how they may affect other tools that use them. Figma has stressed that no Figma content, community files or app designs were used to train the model. It also noted that Make Design is in Beta, while Field has suggested that the problem is that the "variability" of output is too low. He has said that Figma hopes to re-enable the tool soon.
However, some users are saying that the response is far from reassuring. It's hard for users to be confident that they can safely use an AI tool without the risk of legal issues if even Figma doesn't know how the tool was trained and whether it could be directly copying from existing UI designs. Other users are wondering whether its use will lead to a kind of creative convergence, with products looking more similar to each other since they're all copied from the average of existing designs.
Meanwhile, Figma has introduced its own AI training policies, which users have until August 15 to opt in or out of (By default, Starter and Professional plans are in and Organization and Enterprise plans are out, but users can change the default choice.) This suggests the company may be planning to improve Make Design or to develop more AI tools using additional training on users' content.
For more on Figma, see the news about the new Wix Studio Figma plugin and our interview with CEO Dylan Field about Figma UI3.
