Adobe XD adds on-device LLM for secure asset search and copy generation

Design · 4 min read

Adobe announced that XD now includes an optional on-device LLM that runs locally on Mac and Windows machines for asset search, auto-tagging, and short-form copy generation. The model is restricted in size to meet memory constraints and designed to keep user data local, addressing enterprise data residency concerns.

In practice, the on-device model indexes local libraries and design files to enable instant, privacy-preserving search queries like 'find hero images with blue gradients' or 'list buttons missing hover states.' It also drafts UI microcopy in context—error messages, tooltips, and CTA variants—that teams can iterate on without sending data to cloud services.

Adobe will continue supporting cloud-based generative features for teams that opt in, but the on-device option is being marketed to regulated industries and studios with strict confidentiality needs. The company expects the feature to be especially useful for early-stage prototypes and sensitive product concepts.