AI Brain-Computer Interfaces for Accessibility

Brain-computer interfaces (BCIs) read neural signals and convert them into digital commands, bypassing the motor system entirely. For people with severe paralysis, including those with ALS, spinal cord injuries, locked-in syndrome, and advanced motor neuron disease, BCIs represent the possibility of communication and environmental control when no other input method works. AI is the critical enabler: raw neural signals are noisy and complex, and machine learning models decode them into usable commands.

How BCIs Work

Signal Acquisition

BCIs capture brain activity through several methods:

Invasive (intracortical). Electrodes implanted directly in the brain’s motor cortex capture individual neuron firing patterns. Highest signal quality but requires surgery.
Minimally invasive (endovascular). Devices inserted through blood vessels near the brain capture neural signals without open brain surgery. Synchron’s Stentrode uses this approach.
Non-invasive (EEG). Electrodes placed on the scalp capture electrical activity. Lower signal quality but no surgery required.

AI Decoding

Machine learning models translate raw neural signals into intended actions:

The user imagines a movement (moving a cursor up, clicking a button)
Sensors capture the resulting neural activity pattern
AI models trained on that user’s neural data decode the pattern into a specific command
The command is executed (cursor moves, letter is selected, device is controlled)

Training these models requires calibration sessions where the user attempts specific actions while the system learns their neural patterns. Modern AI approaches reduce calibration time and adapt to signal changes over time.

Current State

Neuralink (PRIME Study)

Neuralink’s N1 chip, implanted in the motor cortex through a surgical procedure, reads neural activity at high resolution. As of mid-2025, five individuals with severe paralysis are using Neuralink devices to control digital and physical devices with their thoughts. The company extended its clinical trials to the UK (GB-PRIME study at UCLH and Newcastle), where the first UK patient reported controlling a computer within hours of surgery.

Participants have demonstrated cursor control, typing, and basic device operation. The system is fully implantable and wireless.

Synchron (Stentrode)

Synchron’s Stentrode is inserted via a catheter through the neck into blood vessels near the brain’s motor cortex, avoiding open brain surgery entirely. Synchron raised $200 million to prepare for commercial launch and has demonstrated its BCI controlling home devices and digital interfaces.

In a notable demonstration, a patient with ALS who is paralyzed and unable to speak used the Stentrode with Nvidia AI processing and an Apple Vision Pro headset to control various devices around his home. Synchron’s approach trades some signal resolution for significantly lower surgical risk.

Non-Invasive BCIs

Consumer and research EEG devices (Emotiv, OpenBCI, Muse) provide basic brain-computer interaction without surgery. These systems can detect broad mental states (focused, relaxed) and simple binary commands, but lack the resolution for fine cursor control or rapid typing.

What BCIs Enable for Accessibility

Communication. For locked-in patients who cannot move or speak, BCIs may be the only channel for expressing thoughts. Neural typing speeds are currently modest (approximately 15-30 characters per minute for invasive BCIs) but meaningful for people with no alternative.
Computer access. Web browsing, email, and smartphone control through thought alone.
Environmental control. Operating smart home devices, adjusting beds, controlling wheelchairs, and managing assistive technology.
Prosthetic control. Directing robotic limbs and exoskeletons through neural signals.

Challenges and Limitations

Surgical risk. Invasive BCIs require brain surgery with associated risks of infection, hemorrhage, and device failure. Synchron’s endovascular approach reduces but does not eliminate procedural risk.

Long-term stability. Implanted devices must function reliably for years. Neural signals shift over time as the brain adapts to the implant (a phenomenon called signal drift), requiring AI models to continuously recalibrate.

Speed. Current BCI typing speeds are much slower than any other input method. They are valuable primarily when other methods are unavailable.

Cost. BCI systems including surgery, device, and ongoing support currently cost hundreds of thousands of dollars. Insurance coverage is limited and evolving.

Regulatory pathway. BCIs sit at the intersection of medical device regulation and consumer technology. FDA approval pathways are still being established.

Ethical questions. Neural data is among the most intimate information that can be collected. Who owns it, who can access it, and how it should be protected are open questions.

For alternative input methods, see AI gesture recognition for motor accessibility. For the broader accessibility technology landscape, see the AI accessibility guide.

Key Takeaways

BCIs decode neural signals into digital commands, enabling communication and device control for people with severe paralysis when no other input method works.
Neuralink (invasive, high-resolution) and Synchron (minimally invasive, endovascular) are leading clinical development, with multiple patients using devices in 2025.
AI decoding models are essential: they translate noisy neural signals into usable commands and must continuously adapt to signal changes.
BCIs stand at the cusp of moving from experimental to early clinical use, comparable to where gene therapies were in the 2010s.
Cost, surgical risk, speed limitations, regulatory uncertainty, and neural data privacy are significant barriers to widespread adoption.

Sources

Neuralink — implantable brain-computer interface: https://neuralink.com/
Synchron — minimally invasive endovascular BCI: https://synchron.com/
Willett et al., “A High-Performance Brain-Computer Interface for Typing” — neural decoding for communication: https://doi.org/10.1038/s41586-021-03506-2
FDA BrainGate clinical trials — regulatory information for BCI devices: https://clinicaltrials.gov/search?term=BrainGate