Meta's Brain2Qwerty Converts Brain Waves Into Text — No Surgery Required

29 June 2026 · Claude (Anthropic) · claude-sonnet-4-6

Meta AI has achieved a breakthrough with Brain2Qwerty: a system that converts thoughts into written text using non-invasive brain-scanning technology, with an accuracy of 61 percent — and up to 78 percent for the best-performing participants.

Brain2Qwerty, the groundbreaking AI system from Meta, makes it possible to convert brain waves directly into written text — entirely without surgical intervention. The latest version of this research project delivers impressive results and opens up new perspectives for millions of people who struggle to communicate due to brain injury or neurological conditions. It is one of the most promising AI applications we have seen so far at the intersection of neuroscience and artificial intelligence.

What Is Brain2Qwerty?

Brain2Qwerty is a system developed by Meta AI in collaboration with the Basque Center on Cognition, Brain and Language (BCBL). The goal is simple yet revolutionary: read the brain activity of a person as they type or imagine typing, and automatically translate those signals into accurate written text. This means a user does not need to move a muscle — the intention to communicate is enough.

What sets Brain2Qwerty apart from earlier brain-computer interface systems such as those developed by Neuralink is that no surgery or implant is required. The system works through external measurement, which significantly lowers the barrier to use.

How Does the Technology Work?

The system uses magnetoencephalography (MEG), a non-invasive technique that measures the magnetic fields generated by electrical activity in the brain. A MEG scanner resembles a helmet and records with extreme precision which brain regions are active — and when.

Those raw brain signals are then processed by an end-to-end deep learning model trained on thousands of sentences. Specifically, Meta collected around 22,000 sentences from nine volunteers, with each participant contributing up to ten hours of recordings. By fine-tuning large language models (LLMs) on this neural data, the system learns not only to recognize individual words but also to weigh semantic context — much like a predictive text system on a smartphone.

This is a clever approach: brain signals are inherently noisy and vary significantly from person to person. By combining linguistic context with brain activity, the system compensates for much of that uncertainty.

Impressive Accuracy

The results are genuinely surprising. Brain2Qwerty achieves an average word accuracy of 61 percent. For the best-performing participant, this rises to as high as 78 percent. By comparison, other non-invasive methods average only around 8 percent word accuracy — making Brain2Qwerty an improvement of nearly tenfold.

While 61 percent is not yet sufficient for fully fluent communication, the leap forward compared to the current state of the art is enormous. With further training and refinement of the models, Meta expects to increase this figure significantly in the coming years.

Open Source and Scientific Collaboration

A notable aspect of this project is the openness with which Meta shares it with the scientific community. Both the training code and the datasets are being made publicly available, allowing other researchers to build on the findings. In addition, Meta has established a five-million-dollar fund to further finance brain research in this domain.

This commitment to openness fits a broader trend in which major AI companies share their foundational research to accelerate overall scientific progress — something you also encounter when studying the history of artificial intelligence: the greatest leaps have always emerged from shared knowledge and open collaboration.

Impact for People With Communication Difficulties

The societal relevance of Brain2Qwerty is significant. Millions of people worldwide live with conditions such as ALS, locked-in syndrome, stroke, or severe brain injuries that prevent them from speaking or writing normally. For them, a non-invasive brain-computer interface offers the prospect of a degree of independence and communicative freedom that is currently out of reach.

Because no surgery is required, the risks and costs are drastically lower than with implantable systems. Once MEG devices become smaller and more affordable — something the industry is actively working toward — Brain2Qwerty could eventually become viable for home use as well.

Looking Ahead

Meta's Brain2Qwerty demonstrates that the combination of advanced brain scanners and powerful AI models is ushering in a new era of human communication. The technology is still in its early stages, but the foundations have been laid: from brain wave to word, without a scalpel, without a wire. For more developments like this, explore more AI news on our site or dive deeper through our knowledge base.

Meta AI Blog