On September 14, 2026, researchers at the Weizmann Institute of Science presented Brain-IT, an innovative artificial intelligence model capable of reconstructing images based solely on a person’s recorded brain activity.
The system analyzes functional magnetic resonance imaging, or fMRI, data collected while a person is viewing an image. Brain-IT then recreates the viewed image while preserving important details such as its composition, colors and meaning.
Unlike existing systems, which may require dozens of hours of personalized brain scans, Brain-IT can learn to interpret the brain activity of a new person using approximately one hour of data.
The model identifies patterns of activity shared across different human brains, even when the exact anatomical locations of those patterns vary from person to person. During development, the system identified 128 functional brain regions involved in visual information processing, including several previously unknown functional divisions.
The technology is currently intended primarily for neuroscience research. In the future, similar systems may contribute to the development of advanced brain–computer interfaces and communication tools for patients with paralysis or severe neurological disorders.
Researchers are also exploring whether the technology could eventually decode auditory information, video and possibly even visual content experienced during dreams.
Source: Weizmann Institute of Science