FAIR, Meta

FAIR is Meta’s Fundamental AI Research laboratory, developing models and research tools for machine perception, language and physical-world understanding. Its contributions include Segment Anything, which lets people segment objects in images and videos. The broader model family includes SAM Audio, allowing creators, researchers and developers to isolate sounds using text, visual cues or marked time spans. Users can experiment with their own media in the Segment Anything Playground.

Launched in 2013, FAIR’s founding team was assembled by Mark Zuckerberg and Yann LeCun. Rob Fergus leads its longer-term research within Meta’s reorganized AI operation; LeCun made his departure from Meta to found AMI Labs. The lab’s research direction includes learning through observation: V-JEPA 2 learns predictive representations from video, while its action-conditioned extension uses limited robot interaction data to plan physical actions. Researchers demonstrated picking and placing objects on robot arms without task-specific training.

FAIR shares papers, code, models and demos, with research feeding both Meta’s product teams and outside applications. A concrete example is Alta Daily, which uses Segment Anything to remove backgrounds from wardrobe photographs for its digital styling app. In 2026, Meta reported that Alta had processed more than 20 million images using SAM.