Researchers in Israel demonstrate an AI that can recreate images a person views from fMRI data, offering a glimpse of future medical and communication aids.
A new AI model called Brain-IT can reconstruct images a person sees by analyzing brain activity from MRI scans. It could help people who can’t speak to communicate, but it’s still early and limited to controlled lab settings. In the near term, expect more research into portable brain-signal devices and expanding the types of content AI can decode.
What it means for you: this is a glimpse of how brain data might one day support medical communication, while also highlighting privacy concerns as decoding capabilities improve. What happens next is more experiments, plus talks about safer, smaller tools for real-world use.
The Brain-IT AI model, developed by researchers at the Weizmann Institute of Science in Israel, represents a step toward turning brain activity into visual reconstructions. The team says the model can both reconstruct what a person was looking at and predict how a brain scan would appear if a different image had been shown. In plain terms, it’s not telepathy, but a highly selective pattern-metection trick that translates certain brain signals into pictures.
The key to Brain-IT is pattern recognition. It works by analyzing functional MRI (fMRI) data, which track blood flow and oxygen use in the brain as someone views images. The AI then uses those signals to recreate the scene. The researchers say the model is faster and more accurate at reconstructing both content and details than some earlier methods. They emphasize that this is not a read of thoughts, memories, or language; it’s a structured decoding of visual perception from a specific kind of brain activity.
To train Brain-IT, the team fed thousands of brain scans from eight volunteers who were shown a large set of images. The AI learned to map activity in 128 functional regions of the brain to particular visual features. When a new person was scanned while looking at new images, Brain-IT could reproduce the content with notable fidelity using only about an hour of new fMRI data, far less than prior methods that needed many more hours of data.
The work sits in a cautious middle ground: it’s impressive within a lab, but not ready for everyday use. MRI machines are bulky, expensive, and not something people can casually wear or use at home. The researchers say future work could explore simpler, cheaper devices such as EEG, which measures brain activity with electrodes on the scalp, though EEG generally offers coarser data than fMRI. There’s also interest in extending the approach to audio, with the team hinting at decoding sounds, and even dreams, in the long run.
Right now, Brain-IT is a research project. The practical takeaway is twofold: it highlights how far AI can go in interpreting brain signals for specific tasks, and it underscores potential medical uses. For people who can’t speak or move, a reliable way to translate intended messages into communication could come from brain signals combined with AI. But today you’d still need medical-grade equipment and a clinical setting to get useful results.
The team trained Brain-IT on thousands of brain scans tied to viewing images. They identified 128 regions that light up during perception and built a model that translates that activity into a visual scene. They also showed the model could predict what a brain scan would look like if shown a different image, a useful check on the model’s understanding of the relationship between brain activity and visual content.
There are clear limits. The technology currently relies on fMRI, which is accurate but slow and impractical for everyday use. The researchers are exploring whether EEG could offer a lighter alternative, but that would require solving the challenge of capturing enough detail from scalp signals. They are also aiming to broaden beyond static images to more dynamic content like video or even sounds.
For many readers, this work matters because it spotlights how AI is increasingly used to interpret the human brain. If similar techniques advance, people with severe communication disorders could gain new ways to express themselves. It also raises questions about privacy and who should have access to brain-based data, since such signals could reveal what a person is seeing or thinking in the future.
The researchers say the next steps include testing with simpler, more portable recording devices and extending the AI to other types of brain data. If a path toward at-home or clinical tools begins to emerge, it will likely come from incremental improvements that keep patient safety, privacy, and practicality at the fore.
A new AI model that reconstructs images a person sees from brain scans, using functional MRI data.
No. It decodes brain signals linked to visual perception, not thoughts, memories, or language.
You're reading the quick version.
A freelance designer created the Muse logo for Meta’s AI assistant, then faced sharp criticism online while balancing a small business life.
OpenAI's new video shows three practical Codex-assisted demos that connect telemetry, code changes and security signals to quick fixes in Grafana, Kubernetes and beyond.
A simple guide to AI agents, how they work, and why they’re popping up in workplaces and consumer tools.
The Daily Brief
Every morning, the news that matters from AI, phones, apps and the people shaping tech. Explained in plain English. Free.
One email a day. No spam, unsubscribe anytime. Privacy policy
