In June 2014, in a modest 13-bed hospital in Belize, Phil Kennedy sat before a mirror, staring at his freshly shaved scalp. The Irish-American neuroscientist and inventor, then aged 66, was about to do what many would consider unthinkable, even unethical: subject his brain to an experimental surgery, implanting glass-and-wire electrodes he had developed beneath his scalp.
The modern idea of brain-computer interface (BCI) as a direct communication pathway between the brain and an external device was formally proposed by computer scientist Jacques Vidal in his 1973 landmark paper Toward Direct Brain-Computer Communication. By 1977, he demonstrated the first practical BCI application: a non-invasive electroencephalogram (EEG) system allowing users to navigate a computer cursor through a maze with the mind. Kennedy was, however, part of a small band of pioneers in the 1980s developing ‘invasive’ BCI—electrodes implanted directly in the brain and linked to computers.
Medicine has a long, uneasy tradition of scientists experimenting on themselves, sometimes producing breakthroughs, at times paying a steep price. In 1984, Australian physician Barry Marshall famously drank a broth of bacteria to prove it caused gastritis and peptic ulcers, a gamble that later earned him the Nobel Prize. Kennedy, whose focus was on creating BCIs using electrodes designed to last a lifetime, was drawing from the same daredevil lineage.
To achieve a lifetime design, Kennedy developed neurotrophic electrodes that allow neurons to grow into the implanted material. The patented device consisted of two gold wires housed in a tiny glass cone filled with a proprietary mix of growth factors that stimulate cellular activity. After animal trials in the mid-1990s, the US FDA permitted Kennedy to implant the electrodes in locked-in patients whose paralysis left them unable to speak or move. He supervised at least five such procedures, showing that signals from just a few neurons could let patients move a cursor and communicate by selecting letters or words.
By the late 2000s, however, the FDA withdrew permission for new trials. Also, the inability of the locked-in patients to communicate made it impossible to know what a neuron’s firing represented. So, Kennedy spent nearly a year searching for an ALS volunteer with some vocal ability. When nothing worked, he decided to implant the device in himself.
The surgery was fraught with complications: during the 12-hour procedure, his blood pressure spiked, causing brain swelling and temporary paralysis. When he woke up, Kennedy was initially unable to speak. After recovering from the side effects, he returned to Belize months later for a second operation, during which the surgeon implanted electrodes that allowed Kennedy to record signals from his brain.
Back in the US, Kennedy worked largely alone, recording neural activity as he spoke and imagined sounds, words and phrases. He found consistent neuron patterns linked to speech and imagined speech, but the skull incision failed to heal, forcing implant removal after weeks. The hospital bill came to about $94,000 (then around 57 lakh)—most of it paid out of his pocket.
While Kennedy’s self-funded, high-risk self-implantation marked a dramatic example of the individualistic, ‘heroic’ era of neural exploration—when researchers often faced regulatory hurdles and limited funding—the past decade has seen the BCI sector transform into a high-stakes, corporate and geopolitically driven global race. This shift has been shaped by massive private investment, accelerated clinical trials and aggressive engineering by companies such as Elon Musk’s Neuralink.
Founded in 2016, shortly after Kennedy’s implant, Neuralink helped accelerate the field by refining invasive BCI technology. In January 2024, when the company implanted its first human patient Noland Arbaugh, a quadriplegic—enabling him to control a cursor using thought alone—it echoed Kennedy’s 1990s demonstrations. However, Neuralink achieved far greater precision through 1,024 electrodes distributed across 64 ultra-thin threads.
Where pioneers such as Kennedy prioritised long-term stability and reliable neural recording, often resulting in limited data throughput and modest performance, Neuralink maximised data extraction from the brain. The aim is to enable faster device control, rapid thought-based typing and eventually high-bandwidth integration between human cognition and artificial intelligence.
As of early 2026, Neuralink has implanted devices in at least 21 patients, with research focused on paralysis, speech restoration (which received FDA Breakthrough Device designation in 2025) and vision restoration for the blind. Musk has announced plans for high-volume production starting in 2026, alongside streamlined, robot-assisted surgical procedures that reduce implantation time to under 30 minutes.
Other startups such as Blackrock Neurotech and Synchron have also gained prominence over the past decade. Blackrock focuses on highly invasive, intracortical implants designed for high-resolution neural recording and stimulation, while Synchron emphasises less-invasive, stent-based (endovascular) approaches that avoid open-brain surgery. Even the field of prosthetics is moving in that direction, says Kavinder Beniwal, COO, Motorica India, which specialises in AI-enabled, functional upper-limb prosthetics for adults and children. “Currently, our systems rely more on spinal implants and peripheral nerve stimulation,” he says. “These activate signals through the nervous system to enable movement and feedback. However, all these technologies are converging toward a common goal—direct brain-controlled prosthetics. In the coming years, we expect significant progress in this area. It is entirely possible that within five years, we could see prosthetic hands and legs controlled directly by the brain. That is the long-term vision.”
As ambitions to merge human cognition with artificial intelligence begin to materialise, venture funding in BCIs has surged. “What strikes me most is how quickly the big tech players are rushing into BCIs,” says Velco Dar, entrepreneur and author of the forthcoming book Neuraleap. “Meta is already active with neural wearables. Apple has announced ‘thought as input’ for controlling iOS devices in partnership with Synchron. Sam Altman has launched Merge Labs. And all of this has happened in just the past few months. It is less about the fine print of each project and more about the signal it sends: BCIs are now the space every major tech company wants to be in.” Interestingly, Dar describes consumer wearables such as neural wristbands, Oura rings and smart glasses as “gateway drugs” to BCIs. “They add a novel interface for the consumer but also gather data, build user habits and normalise the idea that our thoughts and signals belong inside the digital ecosystem,” he says.
Arbaugh said that his biggest hope is that BCIs can transform lives of people with disabilities. “I want to see their lives transformed like mine has been,” he says. “I am not sure if it is able to help with ageing, but I absolutely believe it can solve mental health issues, and we haven’t even begun to explore the creative mind yet…. My biggest hope is that BCIs become a tool that restores dignity wherever it is slipping away.”
There are multiple approaches to building BCIs, most commonly distinguished by their level of invasiveness—that is, how directly the technology interfaces with brain tissue. This distinction is crucial, as it shapes everything from signal quality and safety to surgical risk, scalability and potential applications.
Broadly, BCIs fall into three categories: invasive, semi-invasive and non-invasive. At present, however, the industry’s interest lies firmly in invasive systems. Most BCI companies are developing implanted technologies aimed at medical and therapeutic use, and all major players with implantable devices are currently conducting clinical trials exclusively involving patients with disabilities. At the same time, Neuralink’s leadership has publicly signalled a longer-term ambition: extending invasive brain implants to healthy individuals, potentially as early as 2030. Nevertheless, Dar says that non-invasive systems are the obvious bridge from where we are today to a true BCI future.
The first demonstrations of non-invasive EEG-based control of a physical robot date back to the late 1980s. However, it was computer scientist José del R. Millán, in the early 2000s, who pioneered practical non-invasive brain-robot interaction. He showed how human EEG signals could control a mobile robot through a shared control system, making human-robot interaction more intuitive and reliable.
Millán has since developed BCI systems that enable users, including people with severe disabilities, to navigate wheelchairs or telepresence robots using thought alone. His research also demonstrated that users can be trained to modulate their brain signals, a skill that improves over time, and that combining BCIs with rehabilitation can promote neuroplasticity, helping stroke patients and people with paralysis regain function.
Millán predicts that the advanced phase of BCIs will emerge from the convergence of brain signals, artificial intelligence and adaptive robotics—a direction reflected in his own work on brain-controlled robots. He envisions applications beyond mobile robots, including exoskeletons that could restore movement. “If someone’s hand is paralysed but the muscles are still intact, an exoskeleton could move the hand through a BCI,” he explains. “The real challenge is achieving that level of precision with non-invasive EEG.”
This is where shared control becomes crucial. “A BCI can decode only a limited set of commands,” he says, “but when combined with intelligent robotics, those commands can be interpreted in context.” If a user intends to pick up a cup, for instance, the robot can infer how to grasp it based on the object and surroundings. A phone would be handled differently, perhaps from an angle that keeps the screen visible. Sensors and AI fill in these details, much like a baby gradually learning to coordinate hand movements.
Users also need some training—not years, but enough for the AI ‘avatar’ to learn their patterns and intentions. Over time, the brain would communicate only high-level intentions, while the AI executes them efficiently, correcting mistakes through signals known as error-related potentials.
Arbaugh frames the debate between invasive and non-invasive BCIs as a trade-off between precision and convenience. Invasive implants such as Neuralink’s tap directly into the cortex, delivering speed, accuracy and fine control that non-invasive systems cannot yet match. Non-invasive devices, however, score higher on safety, affordability and ease of adoption. In the long run, he argues, both approaches will thrive, but in different lanes. Researchers such as Millán are trying to narrow the gap between the two.
Asked how weak brain signals can be made useful for controlling robots, he says the challenge goes beyond simple amplification. “It is not about making signals louder,” he explains, “but making them better, richer and more informative.” His team is combining neuroengineering, brain stimulation, human learning and machine learning to speed up how quickly users learn to modulate their brain signals.
One approach involves targeted brain stimulation to strengthen key motor regions identified by the BCI, making signals more reliable. Machine-learning models then help stabilise those signals so they remain consistent over time. The next hurdle is complexity. A user may learn to control a single movement—say, operating a robotic exoskeleton—but real life often requires both hands. Signals from a healthy hand can interfere with those controlling the robotic one. Filtering out this interference is now a major focus. “That’s the road ahead,” Millán says— building more robust brain-controlled systems, expanding what they can do, and ultimately making them practical for everyday use by people who need them most.
Notably, BCIs are also emerging as a new front in geopolitical competition. The US currently leads in BCI innovation, but Dar says China is rapidly catching up. “China has announced a state-backed push to accelerate the BCI industry,” he says. “Research hubs, hospitals and industry clusters are being built side by side, creating a truly collaborative ecosystem. What’s striking is that regulation is designed to advance in parallel with innovation. Clinical trials, policy frameworks and commercialisation move together, rather than in isolation. It is a very different playbook from what we have seen elsewhere, and one with clear global implications.”
Dar notes that India, despite its decentralised innovation model, is quietly making serious academic gains. He points to the work at the Indian Institute of Technology Kanpur, where researchers have built a BCI-powered robotic hand using a closed-loop system that synchronises brain signals with a robotic exoskeleton for therapy. “It is a first-of-its-kind device,” he says, “and a clear step toward turning BCI research into real clinical tools.”
But rapid innovation and rising venture capital, Dar cautions, come with heavy responsibility. When private companies handle neural data, privacy concerns are unavoidable. “Neural rights must be protected from the outset,” he says. Chile has already written protections for neural data into its constitution, the EU has classified BCIs as ‘high-risk’ devices, and several US states are drafting laws to get ahead of mass adoption. “Innovation is racing far ahead of regulation. Moving carefully matters more than moving fast, and ethicists and diverse stakeholders need to be part of the process to avoid unintended consequences,” he says.
Arbaugh echoes that warning, arguing that neural data should be treated as the most sensitive medical information imaginable—because it is. Companies, he says, must be open about what data they collect, how long it is stored, who can access it and for what purpose. “No grey areas. Absolute transparency,” he says. While he trusts Neuralink based on direct experience with its standards and people, he adds that trust in one company cannot substitute for clear rules. “Strong regulation protects everyone,” he says, “especially future patients who won’t know the team behind their device.”