Figma Puts Humans First at Config 2026, But Its AI Canvas Runs on Third-Party Tech
Figma doubled down on human-centric design at its annual Config 2026 conference, announcing new tools that keep creative control with users. The catch: the underlying AI models powering its canvas are not built by Figma.
The company revealed a new “Human-in-the-Loop” framework for its generative design features. This means every AI-generated suggestion must be reviewed and approved by a human before it becomes part of a finished prototype. Figma CEO Dylan Field stated the move is to “preserve the designer’s intent, not automate it away.”
Who, What, When, Why: Figma, at Config 2026 (late spring), introduced AI tools that rely on external large language models (LLMs) and image generators. The core reason: rapid experimentation without the cost of proprietary AI research.
“We are not an AI company. We are a design collaboration company that uses AI.” - Dylan Field, Figma CEO, at Config 2026
The AI Is Someone Else’s
Figma confirmed its “Canvas AI” features use models from partners like OpenAI, Anthropic, and Stability AI. The company said this allows it to swap models as technology evolves, avoiding vendor lock-in.
Key details from the keynote:
- OpenAI’s GPT-4o handles text generation and layout suggestions inside the editor.
- Anthropic’s Claude powers code-generation features for developers.
- Stability AI’s SDXL generates image placeholders and background textures.
Figma stressed it does not train its own foundation models. Instead, it builds a thin “orchestration layer” that interprets user prompts and routes them to the best third-party model.
Designers Keep Final Say
The most controversial announcement was the “Auto-Layout Advisor,” which suggests responsive design tweaks. Figma forces users to click “Apply” or “Dismiss” before the suggestion takes effect. No automatic updates.
This stands in contrast to competing tools like Adobe’s Firefly, which can auto-apply changes. Figma’s approach aims to build trust with professional designers who fear AI will replace their jobs.
“We want to make you more productive, not obsolete. Every AI action requires your approval.” - Product VP Yuhan Zhang, on stage at Config 2026
Background: Why Figma Avoids Proprietary AI
Insiders told the press that Figma evaluated building its own image generation model two years ago but shelved the project. The cost of training, hosting, and maintaining a competitive model was deemed too high for a company focused on user interface (UI) tools.
Instead, Figma chose to:
- License existing models through API agreements.
- Build a privacy layer that anonymizes user data before sending it to third-party APIs.
- Offer an on-premise version for enterprise clients that want zero data leaving their infrastructure.
The strategy lets Figma claim “AI-powered” features without the multi-billion-dollar investment that OpenAI or Google makes.
Industry Reaction Divided
Some designers praised the human-first approach. Others criticized Figma for not owning its AI stack, warning that price hikes from third-party providers could eventually be passed to users.
At the same time, Figma’s stock (traded under parent company Adobe) remained stable. Analysts noted the move signals a mature understanding of the design market: users want AI help, but not at the cost of losing creative autonomy.
The full rollout of the new AI features begins in July 2026 for Figma Professional and Enterprise tiers. Free users will get limited access to the GPT-based text tools.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.