Elon Musk’s Neuralink Had a Brain Implant Setback. It May Come Down to Design

Elon Musk’s startup Neuralink revealed that it experienced a problem with its brain implant after the device was installed in its initial participant, 29-year-old quadriplegic Noland Arbaugh.

After the Jan. 28 procedure, Musk was quick to announce on X that Arbaugh was “recovering well” and that “initial results show promising neuron spike detection.” But just weeks later, the company said Wednesday, the device started to malfunction.

Cofounded by Musk, Neuralink is one of several companies developing a brain-computer interface, a system that provides a direct link from the brain to an external device. Neuralink’s technology is designed to decode intended movement signals from the brain to allow paralyzed individuals to move a cursor or type on a keyboard with just their thoughts.

Neuralink’s unique design may have contributed to the device’s mechanical issues. The company’s implant consists of a coin-sized puck that sits in the skull. It holds a battery, processing chip, and other electronics needed to power the system. Attached to this puck are 64 flexible “threads” thinner than a human hair, each containing 16 electrodes. The threads are meant to extend into the brain tissue to collect signals from groups of neurons. But, according to Neuralink, some of those threads didn’t stay in place.

“In the weeks following the surgery, a number of threads retracted from the brain, resulting in a net decrease in the number of effective electrodes,” according to a blog post published by Neuralink. This led to a decline in the rate of data transfer, measured in bits-per-second. A higher bits-per-second value indicates better cursor control.

The blog post doesn’t say how many threads have been pulled out of the brain or how many are still in place. “Suffice to say, that’s not supposed to happen,” says Matt Angle, CEO of Paradromics, an Austin-based company that is also developing a wireless brain implant to help paralyzed people communicate.

Neuralink didn’t offer an explanation as to why or how the threads came out, and the company did not immediately respond to a request for comment from WIRED. However, the brain naturally moves within the skull, and it’s possible that some of the Neuralink threads were pulled out by that movement.

Neuralink and Musk have touted the advantages of its device over the earlier Utah array, a mainstay of brain-computer interface research since the late 1980s. A rigid square, the Utah array contains a bed of 96 tiny shanks, each with an electrode at the tip for recording. Its shanks penetrate the tissue. A major limitation is that it’s not wireless and has to be connected to an external device with a cable.

By contrast, Neuralink’s implant is wireless and has a total of 1,024 electrodes. More electrodes means more data can be collected from the brain—if all those electrodes remain intact.

“Neuralink designed a very novel neural interface,” says Riki Banerjee, chief technology officer at Synchron, a New York company that is currently testing a brain-computer interface in people with paralysis. “They’re learning and that is part of the process.”

Rather than building a device from the ground up, Synchron and Paradromics have taken inspiration from previous medical devices. Paradromics’ design, for instance, is based on the Utah array but makes some key improvements. It’s wireless, for one, and it has 421 electrodes on the end of tiny wires that sit in the brain tissue. Those wires are all much smaller than the shanks of the Utah array, Angle says.

Synchron’s device, meanwhile, is a hollow mesh tube that resembles a heart stent. Instead of going into the brain directly, it’s inserted into the jugular vein at the base of the neck and pushed up against the cortex. Synchron has implanted 10 participants with its device so far, with one surpassing three years with it. (Arbaugh’s implant is still working after 100 days). Banerjee says the company has not seen a decline in signal quality or performance yet.

Andrew Schwartz, a professor of neurobiology at the University of Pittsburgh who builds brain-computer interfaces, also speculates that Neuralink’s design may have caused the implanted threads to push out of the brain.

The brain’s outermost layer of the brain, the dura, needs to be opened in order to place the device. “With multiple wires being inserted individually into the cortex, it may be difficult to suture the dura closed after implanting the wires,” he says. Leaving this opening could have caused scar tissue to form around the opening, leading the threads to withdraw. The Utah array, Schwartz says, is designed so that the dura can be sutured closed after implantation.

Despite Neuralink’s setback, the company still managed to live stream a demonstration of its device on March 20, showing Arbaugh using the implant to play chess just by thinking about it. Arbaugh has also used the device to play the video game Mario Kart. “I just can’t even describe how cool it is to be able to do this,” he said in the video.

In the blog post, Neuralink says it compensated for the lost threads by modifying the recording algorithm to be more sensitive to neural signals. It also says it improved how it translated those signals into cursor movements and enhanced its user interface, and that these changes were able to boost the performance of the device.

For moving a cursor, Angle says having more electrodes doesn’t matter all that much. But for more complex tasks, such as turning text to speech, a higher data rate will be important.

Prior to getting the implant, Arbaugh used a mouth-held stylus, known as a mouth stick, to operate a tablet that had to be put in place by a caregiver. A mouth stick can only be used in an upright position, and it prevents normal speech. When it’s used for long periods of time, it can cause discomfort, muscle fatigue, and pressure sores.

For Arbaugh, Neuralink’s device is “luxury overload,” according to the company’s blog post. He’s still using the implant, which has allowed him to “reconnect with the world” and do things on his own again without needing his family at all hours of the day and night.

“It’s good that the patient can still use the device and he’s still happy with it. At the end of the day, that’s a win,” Angle says. “But from our perspective, companies that are building brain computer interfaces need to be building devices that will be robust and reliable over a multi-year timeframe.”

There are likely to be setbacks on the road toward commercializing brain-computer interfaces, and with Neuralink taking a unique approach with its device, the company could be in for more bumps along the way.