As physicians and neuroscientists, we are concerned about the clinical and scientific limitations of chip-in-the-brain technology. Until these problems are overcome, we argue that less invasive methods could be used.

Implantation of an electronic chip requires opening the skull and inserting the chip into the cortex, where it records spikes from a few dozen neurons (out of millions) for an unpredictable length of time (often less than a year). This technique has been used, for example, to restore motor function after spinal-cord injury in humans (C. E. Bouton et al. Nature 533, 247–250; 2016) and primates (M. Capogrosso et al. Nature 539, 284–288; 2016). In our view, a non-invasive signal could have controlled the movements under study with comparable efficacy.

Many non-invasive brain-derived signals can be used to activate paralysed muscles directly. These signals can be accessed readily throughout the body — for example, by scalp electroencephalograms or by electromyography of muscles. Currently, the risk–benefit ratio of these techniques is more favourable in this context than the invasive approach, particularly when significant functions remain intact post-injury (see go.nature.com/2ohpfmw).

This may change with advances in technology, when the risk–benefit ratio of inserting a chip into the brain improves sufficiently to justify clinically oriented research that can also advance our knowledge of cortical motor control and/or recording technology.