Implanted Brain-Computer Interfaces Carry Hacking and Hardware Risks

Paralyzed patients using implanted brain-computer interfaces reached communication speeds of 78 words per minute in a 2023 study — a five-fold jump from the 15 words per minute reported in 2021. That pace of improvement is the strongest argument for the technology. It is also, according to a recent analysis of the field, the reason clinicians, regulators and patients should now take a hard look at what happens when electrodes sit permanently on the exposed cortical surface and the data they carry flows through networks that others might reach.

How the interfaces work

A brain-computer interface reads electrical signals produced by the brain and translates them into digital signals an external computer can process. The computer then returns instructions — move a cursor, steer a wheelchair, read a sentence aloud — in real time, giving patients with paralysis, stroke or spinal cord injury a channel to act and communicate more independently.

The technology comes in two forms. Non-invasive systems use external electroencephalogram headsets that measure brain activity from the scalp; these already appear in consumer products such as meditation apps and video games. Invasive systems require surgery. Surgeons place electrodes directly onto the exposed surface of the brain — the same electrocorticographic approach reading rooms know from invasive epilepsy monitoring — with the goal of restoring functions such as speech and mobility in people disabled by stroke or spinal cord injury.

The invasive category is where the money and the clinical ambition now sit. Blackrock Neurotech, an early developer; Synchron, an Australian-owned company; and Elon Musk's Neuralink are racing to bring implantable interfaces to patients. Under current regulations, only a handful of clinical trial participants worldwide can access the devices. That scarcity may not last: analysts project the international BCI market will grow from just under A$3 billion today to roughly A$14 billion by 2033.

What the clinical record shows

The World Health Organization counts more than three billion people living with a neurological condition affecting motor, communication or sensory function — stroke, epilepsy, Parkinson's disease, cerebral palsy, traumatic brain injury. For many of them, the communication results are the most concrete.

The 2023 figure of 78 words per minute in paralyzed patients stands as the benchmark, and recent research indicates the technology is still improving quickly.

Beyond communication, the applications map closely onto familiar neurophysiology workflows. Surgeons are using brain-computer interfaces to map brain activity in real time during complex, high-risk procedures where they must preserve eloquent cortex. Sleep researchers are applying the technology to analyze brain signals in patients with suspected insomnia or sleep apnoea, and it may prove more accurate than sleep diaries that depend on participant self-report. Investigators are also testing BCI-based rehabilitation in depression, epilepsy, stroke and Parkinson's disease.

These are investigational and early-clinical uses. No implanted communication BCI discussed here sits in routine, FDA-cleared wide deployment; access today runs through trials.

Three risks that matter

Physical harm. Any brain implant can damage neighboring tissue. Bleeding in a region controlling speech or movement — even a small clot — can impair those functions. Brain infections are rare but can cause swelling and serious complications if not treated immediately. Long-term presence of foreign material inside the skull poses its own problem: over time the brain treats the implant as an intruder, forming scar tissue around it in an attempt to destroy nearby cells and stop the device from working. Routine movements such as breathing can create friction between the rigid implant and soft brain tissue, leaving some regions inflamed.

Cybersecurity. A recent study found that a large-scale breach of BCI systems could theoretically allow hackers to access sensitive neural data, including information described as patients' thoughts and memories. Intrusion could also impair cognitive functions such as concentration, or manipulate motor signals and degrade how well a patient moves. Those scenarios remain theoretical, not demonstrated attacks — but some U.S. jurisdictions, including Colorado through recent legislation, have begun writing neural data rights into law, and major regulatory gaps persist.

Unequal access. An implant currently costs between $50,000 and $140,000, before ongoing maintenance and follow-up care. At that price, ordinary patients are unlikely to reach the technology soon, and the gap between who can and cannot afford it widens as the field advances.

Where the field goes next

Brain-computer interfaces have moved from proof of concept to measurable clinical gains within two years on the communication metric alone. The outstanding questions are durability — what a decade of scar tissue and micromotion does to electrode performance — and governance of the neural data these systems generate. The analysis calls for more high-quality research into the long-term physical and psychological effects of permanent implants, and argues that this work should be publicly funded rather than left to a handful of large, profit-driven companies. How quickly regulators close the gap between trial-grade devices and a A$14 billion market will determine whether the 78-words-per-minute milestone becomes standard care or stays a research benchmark.