begun development

of the AI-enabled brain interface and associated brain-processing semiconductor; it is not yet a completed, clinically deployed chip. The report below reflects that status.

South Korea advances AI brain-to-robot interface technology

South Korean researchers have launched an ambitious brain-computer interface programme aimed at enabling people with severe paralysis to control robotic devices through neural activity while receiving sensory information back from the machines.

The project, led by research teams at the Korea Advanced Institute of Science and Technology, or KAIST, is developing a bidirectional “Brain-to-Robot” platform combining artificial intelligence, neural-interface semiconductors, robotic skin and wearable exoskeleton technology. The system is designed eventually to help people with quadriplegia regain forms of movement and sensation.

Unlike conventional brain-computer interfaces that mainly translate neural signals into commands for computers or other external equipment, the South Korean programme seeks to establish continuous two-way communication between the brain and a robotic device. Neural signals indicating a user’s intention to move would control an exoskeleton, while information detected by the robot would be transmitted back towards the user’s nervous system.

KAIST teams led by Professors Kyoungchul Kong and Jung Kim of the Department of Mechanical Engineering are working with Angel Robotics on the programme, which is being pursued as a flagship initiative under the Korea Medical Device Development Fund. The project began in April 2026 and is scheduled to run until December 2032.

A central component will be a specialised brain chip capable of processing cortical signals generated by the brain. Artificial intelligence will be used to interpret movement intentions and convert neural activity into commands that robotic equipment can execute.

Researchers are also developing encoding algorithms for the reverse journey. Sensors embedded in robotic skin and an exoskeleton would collect information such as touch, ground reaction forces and joint torque. The system would then convert those measurements into signals intended to provide meaningful sensory feedback to the user.

That closed-loop approach distinguishes the project from many existing brain-computer interface systems, which primarily concentrate on decoding neural activity. Companies including Neuralink and Synchron have moved brain-interface technologies into human clinical testing, demonstrating the ability of people with severe paralysis to interact with computers and other devices through neural signals.

Combining motor control and sensory feedback in a single practical rehabilitation system presents a substantially greater technical challenge. Hundreds of channels of cortical activity may have to be processed continuously with extremely low latency if an exoskeleton is to react naturally to a user’s intentions while simultaneously returning sensory information.

The programme could ultimately allow a person with quadriplegia to perform tasks that currently require assistance. Researchers envisage users walking with robotic support, picking up objects and potentially receiving sensations corresponding to contact at their fingertips.

Angel Robotics, founded by Kong, is expected to lead the commercialisation effort. The company has specialised in wearable robotic systems and walking-assistance technology. The development programme is intended to cover the pathway from engineering and clinical evaluation to regulatory approval by the Ministry of Food and Drug Safety and eventual deployment outside laboratories.

The technology nevertheless remains at the development stage. Researchers have emphasised that long-term safety, clinical validation and an appropriate regulatory framework will have to progress alongside the engineering work. Protection of neural data, cybersecurity, risk management and ethical oversight will also be critical if devices capable of interpreting brain activity move into widespread medical use.

KAIST is supporting the project with a broader research programme spanning neuroscience, artificial intelligence, semiconductor engineering and robotics. Professor Sungho Cho’s team is working on AI techniques for interpreting brain signals, while Professor Jihoon Lee’s group is studying low-power neural-interface circuits, wireless neural measurements and on-device AI for closed-loop neuromodulation.

Professor Hyunjoo Lee is pursuing high-resolution neural recording and precision stimulation using miniaturised multimodal electrodes. Professor Minkyu Je is working on AI-based semiconductor integrated circuits for next-generation neural interfaces, while Professor Jae-Woong Jeong’s research includes high-precision brain-signal measurement and neural stimulation.