Meta Unveils Brain2Qwerty v2, AI System That Converts Brain Signals Into Text

Meta has introduced Brain2Qwerty v2, a groundbreaking artificial intelligence system capable of translating brain activity into text in real time using non-invasive brain recordings. The technology represents a major step forward in brain-computer interface (BCI) research and could eventually help people who have lost the ability to communicate due to neurological disorders or brain injuries.

Announced on Monday, the experimental system uses magnetoencephalography (MEG), a helmet-like brain imaging device that records magnetic signals generated by neural activity. Unlike many existing brain-computer interfaces that rely on surgically implanted electrodes, Brain2Qwerty v2 operates entirely without invasive procedures.

According to Meta, the system captures raw brain signals while a person types and processes them using an end-to-end AI model that reconstructs the words and sentences the user intends to write. Rather than depending on manually engineered signal-processing techniques, the model employs deep learning to interpret complex neural activity directly.

To further improve accuracy, Meta integrated large language models into the decoding process. By combining neural data with semantic context, the AI is better equipped to interpret noisy brain recordings and predict the intended text more reliably.

The company revealed that Brain2Qwerty v2 was trained using approximately 22,000 sentences collected from nine volunteer participants, each of whom spent around 10 hours wearing an MEG scanner while actively typing. The extensive dataset enabled the model to learn how specific patterns of brain activity correspond to written language.

Meta reported that the new system achieved an average 61% word-level accuracy, a significant improvement over previous non-invasive brain-to-text technologies, which averaged roughly 8% accuracy. The company also noted that performance continued to improve as more training data was added, suggesting the technology could become even more accurate with larger datasets.

As part of its commitment to open scientific research, Meta announced that it is releasing Brain2Qwerty v2’s source code and dataset through its Digital Brain Project. The initiative also includes a $5 million research fund aimed at supporting the development of open neuroscience datasets and accelerating innovation in brain-computer interface technology.

In an accompanying study published in Nature Neuroscience, Meta researchers explained that although AI has dramatically improved brain-to-text decoding, most of today’s highest-performing systems still depend on surgically implanted electrodes. While these invasive devices often deliver greater precision, they carry surgical risks and long-term maintenance challenges that limit widespread adoption.

Meta believes Brain2Qwerty v2 narrows the performance gap between invasive neuroprosthetics and safer, non-invasive alternatives, potentially making advanced communication technologies more accessible to patients who cannot undergo brain surgery.

The company said its long-term goal is to advance neuroscience by openly sharing research that could contribute to faster diagnosis, treatment, and understanding of neurological conditions.

The announcement comes amid rapid growth in the brain-computer interface sector. Companies such as Neuralink, founded by Elon Musk, and Synchron are developing implantable devices designed to restore communication and mobility for people with severe neurological disorders. Meanwhile, startups including Neurable and AlterEgo are exploring AI-powered, non-invasive technologies that translate brain or neuromuscular signals into digital commands without requiring surgery.

Although Brain2Qwerty v2 remains a research project rather than a commercial product, its latest results demonstrate the growing potential of artificial intelligence to transform how humans interact with computers. If future research continues to improve accuracy and usability, non-invasive brain-to-text systems could open new possibilities for millions of people living with speech and communication impairments.