Scientists in the United States and Japan say artificial intelligence (AI) is bringing humanity closer than ever to translating brain activity directly into words, images and sound, raising extraordinary possibilities for medicine while opening urgent questions about the privacy of the human mind.
At the centre of the advances is a class of technology known as brain-computer interfaces (BCIs), devices that connect the human brain directly to a computer. Combined with machine learning algorithms capable of detecting patterns in complex neural data, researchers say these systems are beginning to achieve what was, until recently, considered impossible.
In one landmark demonstration published in August 2025 at Stanford University in California, a 52-year-old woman who had been paralysed by a stroke nearly two decades earlier watched sentences she could not physically speak appear on a screen in real time. A small array of electrodes implanted in her motor cortex captured the neural signals produced as she imagined speaking, and an AI system decoded those signals into text almost instantly. Researchers described it as the closest science had yet come to practical mind reading.
Months later, a separate research team in Japan reported a system they called “mind captioning,” which used non-invasive brain scans paired with AI to generate detailed descriptions of what a person was seeing or imagining.
At the University of California, Davis, neuroengineer Maitreyee Wairagkar and her colleagues translated attempted speech from a patient living with Amyotrophic Lateral Sclerosis (ALS) into text at roughly 32 words per minute with 97.5 percent accuracy in 2024. Her team is now exploring whether systems can decode inner speech, words formed silently in the mind without any attempt to physically speak.
Frank Willett, co-director of Stanford’s Neural Prosthetics Translational Laboratory, said his team has already detected traces of inner speech passing through the motor cortex during silent mental tasks. “We were able to pick up traces of inner speech pretty clearly in these different tasks,” he said, while cautioning that fully unfiltered inner speech remains beyond current capability.
Researchers have also demonstrated decoding of tone, pitch and rhythm. In a 2025 demonstration, an ALS patient produced variations in pitch and even sang simple melodies through a prototype system.
Separately, scientists are pairing functional magnetic resonance imaging (fMRI) scans with generative AI tools to reconstruct visual images a participant has seen, with experimental efforts also underway to recreate music from brain activity.
Companies including Neuralink, founded by technology entrepreneur Elon Musk, are working to bring implantable brain chips from research settings into commercial use. Wairagkar said the field is approaching a commercialisation threshold. “In the next few years, we will begin to see these technologies being commercialised and deployed at scale,” she said.
Experts caution, however, that as AI grows more powerful and brain implants move closer to mainstream adoption, questions around data privacy, informed consent and the commercialisation of human thought will demand urgent regulatory attention.
