Figma, the popular collaborative design platform, recently announced it is also getting in on the AI hype train. The company’s latest move will also involve leveraging user-generated content to train its AI models, a decision that was somewhat expected to happen.

At its annual conference in San Francisco on June 26, 2024, Figma unveiled “Figma AI,” a suite of AI-powered features designed to enhance user productivity and creativity. These features include automatic layer renaming, design generation from text prompts, AI-enhanced asset search, background removal, language translation, and an AI prototype maker. While these tools promise to streamline workflows and boost creative output, they come with a caveat: Figma intends to use customer content to improve its AI models.

According to Figma’s official statement, sent to users through email, the company needs to “train models that better understand design concepts and patterns, and Figma’s internal formats and structure through Figma content.” This approach aims to develop AI models that are more attuned to Figma-specific concepts and tools, potentially offering users more tailored and efficient AI-driven assistance. Their legal pages have been updated.

The data Figma plans to use for AI training encompasses a wide range of user-generated content. This includes text and images created within Figma, comments and annotations, layer names and properties, and even usage data such as access frequency and technical logs. It’s important to note that Figma distinguishes between “content data” and “usage data,” with the latter being aggregated, de-identified to protect user privacy.

Examples of Usage Data:

• Information on the frequency of content access within your organization.
• Records of technical activities and associated metadata.
• Data collected from telemetry systems.

The company has disclosed that it works with multiple vendors, including OpenAI, regarding the specific AI models and providers powering Figma’s AI features. Figma’s subprocessor list explicitly mentions two key AI service providers: OpenAI, LLC and Jasper AI, Inc. Both are described as “service provider[s] for AI-enabled functionality.”

While Figma emphasizes its commitment to data privacy and security, including encryption of data at rest and in transit, and strict access controls, the prospect of having one’s creative work used for AI training is bound to raise concerns among some users.

We recently saw apprehensions expressed when Adobe, another major player in the design software industry, had modified its Terms of Service that could be interpreted as though the company planned to use user content for AI training.

How to Opt-Out

Prior to Config 2024, we had one setting that managed AI features called Figma and FigJam AI features. For any team and organizations who have this setting disabled as of June 26, 2024, Figma will honor the setting and turn off AI feature access and content training by default.

Figma

Figma has implemented a system that allows users to opt out of content training.

By default, with the one exception quoted above, the content training feature is set to ‘on’ for Starter and Professional plans and ‘off’ for Organization and Enterprise plans. This difference in default settings reflects the typically more complex customer agreements and specific requirements associated with larger organizational accounts.

The process for users who wish to opt out of content training is relatively straightforward. They can navigate to their team settings, find the AI section, and toggle the “Content training” option off. It’s worth noting that this setting won’t take effect until August 15, 2024, giving users six weeks to decide.