Where does Neuralink stand? The company announced, in early 2024, the implantation of its brain interface in a first human patient, able to control a computer cursor with thought alone. This is a real fact, but one to put in context: it is an early medical demonstration in a paralysed person, in the lineage of research published for more than ten years — not the augmentation of a healthy brain, nor a reading of thoughts. The dangers exist and are serious (safety, mental privacy, reversibility), but they are risks to be governed, not science-fiction scenarios.
Definition. A brain-computer interface (BCI) is a device that captures the brain's electrical activity and translates it into a command for a machine (cursor, prosthesis, computer). Neuralink uses an intracortical implant — electrodes placed directly in the cortex — unlike less invasive approaches.
The documented fact is precise: in early 2024, a first human carrying the Neuralink implant was able to move a cursor through the activity of his motor cortex alone. This is a real, heavily publicised engineering advance. But one must read exactly what it establishes. The implant decodes a movement intention — the signal that commands a gesture — and translates it into action on a screen. It does not "read" the content of a mind, does not download memories, does not pilot thought.
This nuance is decisive for untangling the real from the fantasy: decoding a motor command is not reading thoughts. The first is established; the second is not. Confusing the two means leaving the facts for projection. Neuralink is, at this stage, a clinical proof of concept in a few patients, not a consumer product nor a deployed augmentation capability.
This point is essential for judging the "dangers": Neuralink does not operate outside any frame. In May 2023, the U.S. FDA authorised Neuralink to run a human clinical trial, and the PRIME study (Precise Robotically Implanted Brain-Computer Interface) opened for recruitment in the autumn of 2023. This feasibility study, publicly registered on ClinicalTrials.gov, aims first to evaluate the safety of the N1 implant and the R1 surgical robot in people with quadriplegia. In other words, human implantation takes place within an early-phase trial, under regulatory supervision, with a small number of participants.
In May 2025, the FDA also granted Neuralink a "breakthrough device designation" for a speech-restoration application for people who have lost the use of language (ALS, stroke, paralysis). Note the exact meaning of this label: it is a status that accelerates the regulatory review of a device deemed promising — not a marketing authorisation nor a proof of efficacy. That distinction is precisely the kind of nuance that enthusiastic communication tends to erase.
One real, documented danger concerns not mind control but the conduct of the preclinical trials. In December 2022, Reuters reported that Neuralink was the subject of a federal animal-welfare investigation: the Inspector General of the U.S. Department of Agriculture (USDA) was examining possible violations of the Animal Welfare Act, following a complaint filed by the Physicians Committee for Responsible Medicine. According to the elements reported, roughly 1,500 animals are said to have died in the course of the company's work since 2018, and several experiments reportedly had to be repeated after human error.
This file must be held with rigour: these are allegations and an investigation reported by the press and by the complainant organisation, attributed to their sources — not a judicial verdict. The USDA did, at one stage, state it had found no violation during its inspections, and FDA inspectors later noted procedural shortcomings (missing calibration records) during 2023 visits. The honest formulation is therefore: a documented concern and a real investigation, without a definitive conclusion established publicly. It is precisely the boundary between reported fact and proven fault that must be respected.
Brain-computer interfaces have worked, in a medical setting, for more than a decade. Neuralink's advance extends a lineage of published and reproduced research:
In other words, medical BCIs restore lost functions: that is their primary vocation, and it is already changing lives. Neuralink fits into this history — its initial patient is a paralysed person, not a healthy volunteer seeking augmentation.
Neuralink is not alone. The company Synchron advances in parallel with an endovascular approach: its device, the "Stentrode," is inserted via the blood vessels up to the brain, without opening the skull or implanting directly in the cortex. Its U.S. clinical study COMMAND reported positive twelve-month safety results in 2024 in severely paralysed patients. The goal is the same as Neuralink's — to let paralysed people command a computer — but the procedure is less invasive than the intracortical implant.
This plurality of approaches matters: the "Neuralink danger" is not a question of brand, but of technological category. Several companies and laboratories explore different trade-offs between invasiveness, signal precision, and surgical risk. The field is leaving the academic laboratory alone to become an industrial and commercial object — which is precisely what shifts the ethical stakes.
The most useful distinction for judging the dangers is not technical but ethical: it opposes care — restoring a lost function — and augmentation — adding a function to an intact body. To this day, the real use of BCIs, Neuralink included, is therapeutic. Augmentation of the healthy subject remains largely yet to come and hypothetical.
But the trajectory is documented and slippery: the same laboratory, the same techniques pass without a jolt from the therapeutic side to the augmentative one. DARPA indeed launched in 2018 the N3 programme (Next-Generation Nonsurgical Neurotechnology), explicitly aiming at non-surgical interfaces usable by able-bodied people. It is this announced slide — from care to enhancement — that justifies thinking about the dangers now, before they are upon us.
The serious risks of BCIs do not rest on a myth of mind control, but on concrete and already identifiable problems:
None of this is, to this day, an operational capability of "mind control" — it is a risk hypothesis, not a fact. But that is precisely why vigilance is warranted now: one governs before the capability forces access to the inner self.
It is this trajectory that gave rise to the debate on neurorights — the idea of enshrining in law the protection of mental privacy and identity. As early as 2017, researchers led by Rafael Yuste proposed it; Chile legislated in 2021; UNESCO adopted in 2025 a global recommendation on the ethics of neurotechnologies.
The logic is wise: one legislates before the technical capability is fully here, so as not to discover too late that the inner self is no longer protected. Faced with devices like Neuralink, the right reflex is neither science-fiction panic nor advertising wonderment, but the demand for a verifiable framework.
To follow, piece by piece and certainty level by certainty level, the thread that leads from medical research to augmentation and neurorights, the full investigation of Volume 3 — The Soldier of Tomorrow takes up this file with its sources. And to situate Neuralink within the whole subject — from DARPA to the ethics of the augmented soldier — see our reference dossier: military transhumanism, DARPA, and Neuralink.
Does Neuralink read thoughts? No. To this day, the implant decodes a movement intention (the signal that commands a gesture) and translates it into a cursor command. Decoding a motor command is not reading the content of a mind — this second capability is not established.
Is Neuralink dangerous? The real risks are safety (surgery, hacking of a connected device), the privacy of neural data, the reversibility of the implant, and the consent of vulnerable patients. To this is added a documented animal-welfare file during preclinical trials, reported by Reuters. These are issues to be governed, not a mind-control scenario, which is not demonstrated.
What is the difference between Neuralink and Synchron? Neuralink uses an intracortical implant (electrodes placed in the cortex, via open surgery). Synchron employs an endovascular approach: its device is inserted through the blood vessels, without opening the skull, hence less invasive. Both aim first to help paralysed people.
No. To this day, the implant decodes a movement intention (the signal that commands a gesture) and translates it into a cursor command. Decoding a motor command is not reading the content of a mind — this second capability is not established.
The real risks are safety (surgery, hacking of a connected device), the privacy of neural data, the reversibility of the implant, and the consent of vulnerable patients. To this is added a documented animal-welfare file during preclinical trials, reported by Reuters. These are issues to be governed, not a mind-control scenario, which is not demonstrated.
Neuralink uses an intracortical implant (electrodes placed in the cortex, via open surgery). Synchron employs an endovascular approach: its device is inserted through the blood vessels, without opening the skull, hence less invasive. Both aim first to help paralysed people.
Dossier : Transhumanisme militaire & DARPA
Accueil · Collection · Journal · Dossiers · Sources