The Question
The experiment was precise and limited. Rajesh Rao, a computational neuroscientist, wore an EEG cap recording his scalp's electrical activity as he watched a video game. When he imagined pressing the fire button, his brainwave pattern changed characteristically. That pattern was transmitted over the internet to a transcranial magnetic stimulation coil over the motor cortex of Andrea Stocco, a colleague in a different building on campus. The coil fired, and Stocco's right hand moved in a twitching, involuntary motion, as if a puppeteer had briefly taken control. He had no idea when it would happen and could not stop it — his hand moved because someone else thought it should.
This was not rich mental communication. Rao transmitted no thought, image, or concept — only a binary signal, fire or don't fire, decoded by software into a magnetic pulse. It took an EEG cap, a computer, the internet, and a stimulation coil. The bandwidth was roughly one bit. But the principle — that signals recorded from one brain can be sent across a network and used to influence another — was genuine, and the team repeated it with more participants in 2014.
What the Evidence Shows
The Rao–Stocco experiment was not the first brain-coupling demonstration. Earlier in 2013, Miguel Nicolelis's team at Duke University linked the brains of two rats through implanted electrodes — one in North Carolina, one in Brazil — so that signals from one helped the other solve a task. In 2015, his group reported networks of linked rat and monkey brains, which Nicolelis called a "brainet".
The 2019 BrainNet experiment at the University of Washington took the concept further. Three people played a Tetris-like game: two "senders" could see the whole screen and decided whether a falling block should be rotated, and their decisions were sent to a "receiver" as flashes of light induced by magnetic stimulation. Five groups played 16 rounds each and succeeded, on average, 81% of the time. It was slow and crude — but it was the first multi-person brain-to-brain interface.
"It was both exciting and eerie to watch an imagined action from my brain get translated into actual action by another brain."
— Rajesh Rao, University of Washington, August 2013Separately, decoding the brain with AI has advanced dramatically — mostly with surgical implants. In 2021, Edward Chang's team at UC San Francisco decoded attempted speech from a paralysed man's brain at up to 18 words per minute; in 2023, teams at UCSF and Stanford reported implant-based speech decoding at about 60 to 80 words per minute. Also in 2023, a University of Texas team used fMRI scans to reconstruct the gist of language a person heard or imagined, and Meta researchers decoded rough images from brain scans. These are brain-to-computer systems, not brain-to-brain links, and the non-invasive ones require large scanners and hours of training data per person.
We are not yet at telepathy. We are at the telegraph stage of brain-to-brain communication — one bit at a time, with enormous infrastructure. But the telegraph became the telephone, and then the internet.
Why This Is Happening
Neural decoding has crossed a threshold of practical utility. For decades the main obstacle was not stimulating or recording from a brain — both were achievable — but decoding the signals into something meaningful. Deep learning changed that. Neural networks trained on brain imaging paired with stimulus or behaviour data now decode signals with accuracy impossible using traditional signal processing — the same AI behind image recognition and language models, applied to the brain.
Non-invasive technologies are improving, slowly. Early experiments relied on EEG, which captures only crude, averaged signals through the skull. AI signal processing and better hardware have improved what can be extracted from scalp recordings, and wearable functional near-infrared spectroscopy (fNIRS) is advancing. But the skull blurs signals, and the highest-quality decoding still requires surgery or room-sized machines.
Military research has funded the field. DARPA's Next-Generation Nonsurgical Neurotechnology (N3) programme, launched in 2018, funded research into non-invasive or minimally invasive interfaces that can read from and write to the brain. Such programmes have underwritten much of the basic science, although no non-invasive device has yet come close to the bandwidth of implants.
What Could Happen
Building on the implant-based speech decoding at UC San Francisco and Stanford, systems that let paralysed patients communicate through decoded neural signals become approved medical devices. The "receiver" is a screen or synthetic voice rather than another brain — but it is direct communication from a mind in a meaningful sense, and it has the clearest clinical justification.
A wearable EEG device, a powerful AI decoding layer, and a paired stimulation device worn by the receiver let simple signals — emotional valence, yes/no decisions, directional commands — pass between two consenting users at useful speed. It is marketed for gaming, intimate relationships, or team sports, with early clinical uses for non-verbal communication in autism or locked-in conditions. Bandwidth is still very low by language standards, but the symbolic importance is huge.
Invasive high-density recording (Neuralink, BrainGate), AI decoding of language and imagery (the UT Austin and Meta teams), and minimally invasive stimulation combine to transmit complex mental content — words, images, concepts — directly from one brain to another with enough fidelity to constitute genuine mind-to-mind communication. This requires solving several hard problems at once — less likely within the timeframe, but not impossible.
What Can We Do
Transmitting mental states between people — even at low bandwidth — raises questions about consent, privacy, and identity that philosophy and law are unprepared to answer. The time to prepare is now, while the technology is still crude enough that its implications can be engaged thoughtfully.
Establish the legal principle of cognitive liberty before it is needed. The right to mental privacy — not to have your thoughts read, transmitted, or influenced without consent — is starting to enter law. Chile amended its constitution in 2021 to protect "mental integrity" and brain activity, and in 2024 Colorado and California extended their privacy laws to cover neural data. Scholars including Marcello Ienca and Roberto Andorno have argued for a human right to cognitive liberty. Establishing it before these technologies deploy commercially is far easier than retrofitting it later.
Require informed consent standards specific to neural communication. Consent to have your neural data recorded and transmitted to another person's brain is qualitatively different from sharing a photograph or medical record. It needs its own framework — specifying what is recorded, how long it is retained, whether it can train AI models, who can access it, and what happens if it is hacked or misused.
Fund research into the security vulnerabilities of neural devices. Transcranial magnetic stimulation can make a muscle twitch, as the 2013 experiment showed, though it requires bulky equipment and a willing participant. As neural implants and wearables spread, security research into how they could be hacked or misused deserves more attention and funding.
Begin public deliberation about networked-brain ethics before it is needed. BrainNet showed that people can share simple decisions brain to brain. Deliberation about whether and when such links would be acceptable — in workplaces, classrooms, or the military — is best done while the technology is still crude.
- University of Washington — Researcher controls colleague's motions in first human brain-to-brain interface, August 2013
- University of Washington — BrainNet: play a video game using only your mind, July 2019
- Pais-Vieira M et al. — A brain-to-brain interface for real-time sharing of sensorimotor information, Scientific Reports, 2013
- Moses DA et al. — Neuroprosthesis for decoding speech in a paralyzed person with anarthria, New England Journal of Medicine, 2021
- Tang J et al. — Semantic reconstruction of continuous language from non-invasive brain recordings, Nature Neuroscience, 2023
- Ienca M & Andorno R — Towards new human rights in the age of neuroscience and neurotechnology, 2017