A brain-computer interface (BCI) captures the electrical activity produced by neurons, translates it into digital commands, and then drives a device — a cursor, a prosthesis, a robotic arm. The principle is documented and has worked for more than a decade in a medical context: in 2012, the BrainGate trial published in Nature (Hochberg et al.) showed people with tetraplegia commanding a robotic arm through the activity of their motor cortex alone. But one decisive point structures the entire subject: a BCI reads a motor intention — the signal that commands a movement — it does not read the content of a thought.
Definition. A BCI is a system that establishes a direct communication channel between the brain and a machine, bypassing the nerves and the muscles. It is always made of three links: a sensor that records neuronal activity, a decoder that interprets that signal, and an effector (the commanded device) — often completed by feedback to the user.
Whatever the device, the operation of a brain-computer interface follows the same chain, from neuron to machine:
What makes this possible is that the intention to move leaves a readable electrical signature in the motor cortex, even in a paralysed person whose muscles no longer respond. The BCI short-circuits the broken nerve pathway: it wires the projected gesture directly to the machine.
BCIs split along one major dividing line: must the skull be opened, or not? On that choice depend the precision of the signal and the level of risk.
Closing this gap — the precision of the implant without the surgical act — is precisely the goal set by DARPA, the Pentagon's advanced research agency, with its N3 programme (Next-Generation Nonsurgical Neurotechnology), launched in 2018-2019. N3 explicitly targets non-surgical brain-machine interfaces usable by able-bodied people, including for defence applications. This is the threshold where the BCI ceases to aim only at medical repair and begins to explore the augmentation of the healthy subject.
This is the most misunderstood distinction in the field, and the most important one for staying honest. Reading a motor intention is an established and reproduced reality: decoding the signal that commands a hand movement is today a scientific fact. Reading thoughts, on the other hand — accessing the content of a mind, ideas, memories or inner language — is not an established capability.
Decoding the signal that commands a gesture is not reading the content of a mind. A motor BCI recognises an activation pattern tied to a projected movement; it no more "reads" thought than a thermometer reads an emotion. The most spectacular announcements — "brain-to-brain" communication, synthetic telepathy — belong to exploratory research ambitions, not to attained capabilities. As early as 2009 DARPA had funded a programme named Silent Talk to that end: its public results remained limited, which illustrates the gap between the stated objective and reality.
The BCI has left the purely academic frame. DARPA's Revolutionizing Prosthetics programme led to the advanced "LUKE" prosthetic arm (DEKA), authorised by the US FDA in 2014. On the private side, the company Neuralink announced in early 2024 the implantation of its interface in a first human patient, able to command a cursor by thought — in the precise sense, here again, of a decoded motor intention. Other companies are advancing in parallel, such as Synchron, whose endovascular approach (a sensor slipped into a blood vessel of the brain) avoids open cranial surgery.
This trajectory, from the therapeutic toward the potentially augmentative, is exactly what gave rise to the debate on neurorights — the legal protection of mental privacy. An interface that reads motor intention could, in principle, read more; better to debate it before the technical capability is here. None of this is today a "mind-control" capability: it is a hypothesis of risk, addressed preventively, and not a fact.
DOCUMENTED FACT. On 20 May 2019, DARPA announced funding for six teams under its Next-Generation Nonsurgical Neurotechnology (N3) program, launched in March 2018: Battelle Memorial Institute, Carnegie Mellon University, the Johns Hopkins Applied Physics Laboratory, PARC, Rice University and Teledyne Scientific. The stated goal: wearable, bidirectional, high-resolution brain-machine interfaces for able-bodied service members, with no surgery.
DOCUMENTED FACT. The four-year program runs on two tracks: "completely noninvasive" fully external interfaces, and "minutely invasive" systems using nanotransducers delivered temporarily and nonsurgically. Approaches combine optics, acoustics and electromagnetics. The explicit technical target: communicating with the brain from a range of "just a few millimeters." Program manager Al Emondi described a future headset a soldier would put on for a mission, then set aside.
CORRELATION. N3 continues two decades of DARPA military neurotech — neural control of prosthetics (Revolutionizing Prosthetics), restored touch, memory. The national-security applications DARPA itself cites are explicit: active cyber-defense systems, drone swarms, human-machine teaming on complex missions.
WHAT N3 DOES NOT PROVE. A research program describes goals, not demonstrated capabilities. The 2019 announcement funds teams and sets ambitious targets — it does not establish that an operational "thought-reading" headset exists, nor that the goals were met. "Writing" to the brain means localized neuromodulation, not implanting thoughts or memories. Mistaking a program's stated ambition for a delivered result is the most common analytical error on this topic.
Can a brain-computer interface read my thoughts? No, not in the documented state of the technology. Established BCIs decode a motor intention — the signal that commands a movement — which is very different from accessing the content of a mind. Reading thoughts in the ordinary sense is not an established capability.
Do you need an implant in the brain for a BCI to work? Not always. Invasive interfaces (intracortical implants, like BrainGate or Neuralink) offer a very precise signal but require surgery. Non-invasive interfaces (EEG on the scalp) avoid the operation but capture a blurrier signal. DARPA's N3 programme seeks to combine fineness with the absence of surgery.
Do BCIs really work, or is it science fiction? They really work, in a medical context, and have for more than a decade: the BrainGate trial (Nature, 2012) showed people with tetraplegia commanding a robotic arm. What still belongs to speculation is the reading of thoughts or "synthetic telepathy", not to be confused with the demonstrated capabilities.
To follow, piece by piece and certainty level by certainty level, how the BCI moves from care to augmentation — and why this engages the sovereignty of consciousness — our full investigation lays out all the primary references: the ebook The Soldier of Tomorrow. For the overall panorama of military transhumanism, from DARPA to Neuralink, see also our pillar article.
No, not in the documented state of the technology. Established BCIs decode a motor intention — the signal that commands a movement — which is very different from accessing the content of a mind. Reading thoughts in the ordinary sense is not an established capability.
Not always. Invasive interfaces (intracortical implants, like BrainGate or Neuralink) offer a very precise signal but require surgery. Non-invasive interfaces (EEG on the scalp) avoid the operation but capture a blurrier signal. DARPA's N3 programme seeks to combine fineness with the absence of surgery.
They really work, in a medical context, and have for more than a decade: the BrainGate trial (Nature, 2012) showed people with tetraplegia commanding a robotic arm. What still belongs to speculation is the reading of thoughts or "synthetic telepathy", not to be confused with the demonstrated capabilities.
No. N3 (announced in 2019, over four years) funds six teams to develop non-surgical brain-computer interfaces reading and modulating neuronal activity to within a few millimetres. "Writing" means neuromodulating neurons, not implanting thoughts; and these are research goals, not a deployed device. Source: DARPA, 20 May 2019.
Dossier : Transhumanisme militaire & DARPA
Accueil · Collection · Journal · Dossiers · Sources