Electrodes that can turn brain signals into speech and movement

With up to a hundred metal spikes on its square-shaped surface, the Utah array looks as if someone decided stepping on a Lego wasn’t painful enough.

Thankfully, the Utah array is about half the size of a fingernail. It’s also not supposed to go anywhere near the floor. Instead, while the rest of the tiny device stays outside the brain, those small spikes go into the brain like a pushpin, where they serve as information-gathering electrodes as part of an interface between the soft, organic body and digital machines.

Brain-computer interfaces (BCIs) are setups that connect the human brain to machinery. The devices allow users to send brain signals to a computer, potentially moving a cursor, texting, or creating automated speech. They are a vital part of adaptive technologies that perform tasks a person cannot do on their own, such as speaking or moving.

For people who are severely paralyzed because of an injury or a condition like amyotrophic lateral sclerosis (ALS), a BCI can be life changing. In June of this year, researchers at the University of California, Davis, and colleagues from the research collective BrainGate reported that a man with ALS who was fitted with a BCI used the technology to communicate independently for almost 2 years (Nat. Med., DOI: 10.1038/s41591-026-04414-6).

“You have a man who’s paralyzed. He can’t move his arms and his legs. He can’t be understood when he speaks. And using this technology, he’s gone back to work full-time,” says David Brandman, the principal investigator on the study. “Using this technology, he can have a conversation with a 6-year-old daughter.”

When talking about BCIs, the Elon Musk–founded company Neuralink immediately comes to mind. But Neuralink is just one of multiple international companies building these interfaces and jockeying for regulatory approval. A Neuralink founder, Ben Rapoport, left the company to create Precision Neuroscience, where he focuses on developing surface electrodes that don’t damage the brain. Synchron is adapting stent technology to make insertion easier. Across the BCI space, multiple start-ups and research groups are leveraging different materials and strategies to bring potentially life-changing BCIs to people living with various conditions.

How does a brain-computer interface work?

BCIs listen in on the brain’s activity and relay those electrical signals to a computer, where they can be decoded into actions.

The process starts with measuring the relative charge differences between neurons. All your thoughts and your senses are driven by electrical signals called action potentials. As ion channels open and close in nerve cells and charged ions flow in and out of the neuron, an electrode can pick up signals from the changes in electrical potential.

The resulting readings are like threads that can be stitched together to create something much larger. Doing so takes a very, very powerful weaving program that can learn as it goes. But if it’s successful, the electronic signals can be interpreted by a computer like instructions: move a cursor this way, type out a message, make this sound. When it all comes together, a debilitating disease like ALS may become just a bit more manageable.

One of the easier ways to measure these differences in charge—to gather threads—is with an electroencephalogram, or EEG, in which electrodes are placed on the scalp to detect the differences in charge. This method is the least invasive, and it’s useful for measuring brain activity, but it can’t pick up signals from very deep in the brain because they’re blocked by the scalp and skull.

For a BCI, most electrodes need to be much closer to the brain, but the base concept of how they work does not differ greatly from that of electrodes outside the brain.

In the case of BCIs, says Cindy Chestek, a BCI researcher at the University of Michigan, the trick is implanting the electrodes in the right place. While they generally don’t need to be positioned on the exact neurons, the closer you are, the stronger the signal is.

Once the electrode is implanted in the right spot and there are enough neurons around, decoding the signals it carries comes down to machine learning.

What’s the best material for a brain electrode?

No matter where in the brain the electrode goes, it’s necessary to consider how the body will react to the foreign material and which material is best to use. The materials used in these electrodes often determine how much scarring forms in the brain tissue: while silicon is a common electrical engineering material, it prompts an immune response from the brain. This response causes scar tissue formation that may block neuronal signals.

Some researchers, such as Chestek, are developing carbon-fiber electrodes tipped with platinum. Chestek says carbon fiber is likely as close as we can get to true biocompatibility—it doesn’t fracture like silicon or glass, and it doesn’t deform like metal.

Inbrain Neuroelectronics is a BCI start-up that uses graphene for its electrodes. The Barcelona, Spain-based firm is developing electrodes that listen in on brain activity, and it is also making BCIs that modulate brain activity, which could be a potential treatment for Parkinson’s disease. This modulation is done by electric stimulation, which requires a brain-penetrating electrode. Inbrain uses graphene for its electrodes not just because of its biocompatibility but also because of its ability to withstand the brain’s electrical activity.

“Graphene has a superpower, which is the charge injection limit, so the capacity to stimulate without degrading the material, without creating redox reactions,” says Carolina Aguilar, Inbrain’s CEO and cofounder. Essentially, graphene doesn’t degrade as much in the brain as a metal like platinum would.

Then there are the small, spiky Utah arrays, like the one used by the research collective BrainGate, led by the UC Davis researchers who reported success in enabling a person with ALS to communicate independently. These arrays are clusters of millimeter-sized electrode needles that are tipped with either platinum or sputtered iridium oxide, which conducts the signal.

How deep does an electrode need to go?

Even if BCI electrodes don’t need to be directly attached to the precise neurons responsible for a particular signal, they do need to be closer to the brain than an EEG can get to pick up the signals. And the necessary depth of implantation is debatable.

Some companies, such as Precision Neuroscience and Kampto Neurotech, are developing flexible electrodes that can sit on top of the brain and record signals through electrocorticography. They claim that this will be close enough to get enough of a signal to form a BCI.

“I think you can do a lot from the surface,” says Kenneth Shepard, Kampto’s founder and acting CEO. He explains that he believes there is enough information in local field potentials, or the charges around the cells, at the surface of the brain, referencing a 2025 Nature Electronics paper from his group at Columbia University (DOI: 10.1038/s41928-025-01509-9).

Kampto’s BCI, called Biological Interface System to Cortex or BISC, sits on the brain’s surface and is made of silicon, a common material for computer chips. Silicone is usually hard and rigid, but the BISC electrode is so thin that it can conform to the shape of the brain.

“We thinned the chip down so much it basically sticks to the brain surface, and it’s like a piece of wet tissue paper on a bowl of Jell-O. As the Jell-O moves around, the tissue paper on the Jell-O surface moves around with it, so there’s no relative motion,” Shepard says. Reducing motion reduces the foreign-body response, he says, in contrast to a fixed point that generates friction and stress around itself as the brain moves around.

The New York firm Precision, which is also developing BCIs that sit on the brain’s surface, uses typical conductive metals like platinum for the conductor part of its Layer 7 Cortical Interface electrode, which is then insulated with polyamide. Similar to Kampto’s approach, placing the electrode on the surface without penetrating the brain typically evades the foreign-body response, says Vanessa Tolosa, Precision’s senior vice president of research and development operations.

“It’s almost like we have this cheat code,” she says.

The company Synchron, which is based in Brooklyn, avoids touching the brain at all. Instead, its device, the Stentrode, builds on existing stent technology. A stent is a mesh tube that opens blocked arteries throughout the body and keeps them open; using this device is a common intervention for heart disease.

Kurt Haggstrom, Synchron’s chief commercial officer, explains that the Stentrode is placed in the superior sagittal sinus, a blood vessel that sits atop the brain. From there it picks up neuronal signals and transmits them to a transceiver implanted in the chest.

“Over time, what happens is that the stent gets incorporated into the blood vessel through a process called endothelialization, and essentially what that does is turn your blood vessel into an antenna,” Haggstrom says. Synchron’s bet is like that of Kampto and Precision: that there is enough information at the surface of the brain to make something useful out of it. But Chestek has doubts that these surface electrodes will be able to get enough data, even with the help of artificial intelligence, because they can’t get the activity of single neurons from there. She likened it to trying to turn your cell phone camera into a telescope: even if AI could perfectly analyze every speck of data in that photo, the camera just isn’t good enough to get an image of something really far away.

Chestek and Brandman both agree that surface electrodes can’t get the activity of individual neurons. Brandman notes that different BCI applications could have different requirements for how close the electrode is. Some people may benefit more from a surface electrode, while others may benefit more from a penetrating electrode. But no one knows what method will be best for each person yet.

“We’ve got to study them,” Brandman says.

What comes next for BCIs

There is no BCI approved by the US Food and Drug Administration as of publication. All these programs and devices are investigational, but progress is being made.

Chestek says the first approvals will likely come for BCIs that decode speech, like the one from BrainGate. “I think the first thing you’ll see is speech decoding, and it wouldn’t surprise me if somebody got approval for that in like 3 years,” she says. “Just because going from not being able to speak at all to speaking at a decent speed through a speaker is enough to jump over the line.”

Precision, Synchron, Kampto, and Inbrain have all tested their devices in humans: Precision says it has data from almost 100 patients, 15, Kampto from 4, and Inbrain from 8 people over a short surgery time, after which the devices were removed.

Precision’s Tolosa says she hopes the minimally invasive approach can make a BCI more accessible and less intimidating for people who could benefit, which will continue to push the field forward. A less involved process enables more people to get these experimental BCIs, which means more data that this method works, which eventually leads to lowered barriers for access, she says. “But I think we need big wins to show the world what this technology can do.”