Science · Health
Musk has spent years promising to restore sight. A Spanish team is already testing it in blind people (and the man who leads it urges caution)
In June 2025 Neuralink promised its first vision implants within six to twelve months. A Spanish team has already made blind people perceive points of light and simple shapes with temporary implants, and in China the first commercial brain-computer interface has left the laboratory.
In June 2025, at Y Combinator, Elon Musk said Neuralink would carry out its first vision implants within six to twelve months. It is what he has been repeating since 2024: that his Blindsight implant will restore sight. It is his favourite promise. It is also, so far, the one with the least evidence behind it. As of 4 November 2026, Neuralink has not implanted Blindsight in a single person, nor published a single peer-reviewed study on vision. Meanwhile, in the Spanish city of Elche, two blind people have perceived points of light with an implant they wore for six months, and before them a third came to identify letters.
If we are talking about brain implants for sight at all, it is largely thanks to Musk. None of the other teams here has his megaphone; the field would probably have moved forward without Neuralink anyway, but quietly, with far less money and attention. The question is not whether Musk helps the cause (he does, a lot), but separating what is already real from what is still a promise.
Musk’s promise is the way into a story almost nobody is telling in the right order: brain implants have stopped being science fiction, but the ones that actually work today do not carry his name. A team at the Universidad Miguel Hernández, a public university in Elche, in south-eastern Spain, published its results in November 2025; in China, the first invasive brain-computer interface cleared for commercial use is already being fitted in hospitals; and one very recent precedent explains why any patient should look at these promises calmly.
The promise, dated
Musk’s history with blindness is a run of deadlines that keep moving. In September 2024 he announced on X that Blindsight would let people see “even those who were born blind”, so long as the visual cortex is intact. That same month the FDA granted it Breakthrough Device designation, which speeds up the paperwork but is not an approval. In May 2025, at the Qatar Economic Forum, he put the first patient at “the end of this year or the start of next”. A month later, at Y Combinator, he talked about “six to twelve months” for the first vision implants and warned that the resolution would be quite low at first. In January 2026 he left it “awaiting regulatory approval”. The calendar has changed several times; the promise has not.
The promise vs. the reality (2024–2026)
Every date is a fresh promise about Blindsight. The reality, below, has not moved. Tap any milestone to see how much time has passed.
-
17 Sep 2024
“You will see even if you were born blind”
Said on X. That month the FDA grants Blindsight Breakthrough Device designation (it speeds up the paperwork, it is not an approval).
-
20 May 2025
First patient “in late 2025 or early 2026”
At the Qatar Economic Forum.
-
16 Jun 2025
“6–12 months” to the first vision implant
At Y Combinator. He warns that the resolution will be quite low at first.
-
31 Dec 2025
Large-scale production promised for 2026
On X.
-
28 Jan 2026
“Awaiting regulatory approval”
On X. That day, Neuralink counts 21 participants on its website — motor control, not vision.
Source: Musk’s public statements on X and at Y Combinator and Qatar Economic Forum events — a primary but self-interested source, with no peer-reviewed study and no ClinicalTrials.gov registration behind them; Neuralink, 28 Jan 2026 (participants) and 17 Sep 2024 (announcement of the FDA designation). The count in the band was checked on 4 Nov 2026 on ClinicalTrials.gov, PubMed and Neuralink’s website.
What is actually true at Neuralink
Separate what has been demonstrated from what has been announced. Neuralink had 21 participants in its trials worldwide in January 2026 (the company said so on its website on the 28th of that month), and by May 2026 it was on its 26th patient: on 20 May police sergeant Lee Marten was implanted at Toronto Western Hospital, the first Canadian ALS patient, but those trials are about controlling devices with thought: moving a cursor, operating a computer. They are not about vision. Blindsight, the implant for sight, has no human patient yet, is not registered on ClinicalTrials.gov, and there is no peer-reviewed publication from Neuralink on visual restoration. Everything known about Blindsight comes from its founder and from the company itself. That does not invalidate it, far from it, but it puts it where it belongs: in a very ambitious, hopeful promise, not in the evidence.
The turn: what is happening far from Neuralink
In November 2025, a team at the UMH in Elche, led by Professor Eduardo Fernández Jover with the neurosurgeon Pablo González López (of the Dr Balmis Hospital in Alicante), published in Science Advances the results of implanting a 100-electrode array in the visual cortex of two blind people, who wore it for six months and then had it taken out in a second operation. According to the university’s press release, the participants came to perceive patterns and movement. The letters come from a different experiment and a different person: the same team had published them in 2021 in the Journal of Clinical Investigation, with 96 electrodes and a single participant, a 57-year-old woman, who came to identify some letters and the edges of objects. They are two separate studies. This is not natural vision (they are phosphenes, points of light the brain learns to interpret) and it is still research, not a device a patient can ask for.
What that study adds is not how many points of light can be produced but how the work is done. The implant does not only stimulate: it also records how the neurons respond, and the paper puts forward, as a future step, adjusting the next stimulation to that response. “You have to keep up a dialogue with the brain,” Fernández explains in a phone conversation on 22 September 2026.
And there is one case that goes further, which his own team treats with caution. In February 2026 it published in Brain Communications the case of a participant blinded by ischaemic optic neuropathy in both eyes, whose implant was removed after six months of experiments. Eighteen months after its removal, the participant kept an improvement nobody had expected: perceiving light and movement again, and reading large letters and words. The improvement was in his own natural sight, not in the artificial vision the implant produces; it declined over time after the implant was taken out, and he remains within the WHO blindness range. It is a single case, and the paper’s title sums it up: The unexpected sight.
The classic model: what a patient with a cortical implant would see
Compare a real scene with its phosphene simulation. Slide to reveal, and change the number of electrodes.
An average human retina has 92 million rods and 4.6 million cones (Curcio et al., 1990). Today’s cortical implants run from dozens — Orion, 60 electrodes — to around 1,000. The simulation follows the model the classic studies start from: one electrode, one point of light. Fernández sees it as a good theoretical basis, but it is not what his patients describe: “They don’t see points as such, but shapes.” Sometimes something defined, like a straight line or a cross; at other times something diffuse, like a cloud. Source: UMH Elche (Science Advances); Second Sight Orion; Eduardo Fernández, phone conversation, 22 September 2026. Illustrative simulation.
The second focus is China. On 13 March 2026, the Chinese regulator (NMPA) approved the NEO implant (developed with Tsinghua University) as the world’s first invasive brain-computer interface for commercial use. Mind the caveat: NEO does not restore sight; it is designed for people with tetraplegia from a cervical spinal cord injury, and it helps them recover movement in the hand. Its relevance to this story lies elsewhere: a brain implant has left the clinical trial and is already being fitted in ordinary hospitals, which has yet to happen in the United States or the European Union.
Brain implants, by stage of development
Seven active projects, one abandoned. Filter by stage, or open each group for the detail.
Commercially approved
Clinical trial
Preclinical
Abandoned
Precision Neuroscience (US, temporary use: 510(k) K242618, March 2025, under 30 days) has no stage comparable to the ones above: it has no trial registered on ClinicalTrials.gov, and its 510(k) was cleared with no clinical data, only bench testing and animal work. Blackrock Neurotech has no trials registered in its name either. The longest trial of a brain implant of this kind, BrainGate, documents 14 people implanted between 2004 and 2021.
Source: Neuralink; Reuters; NMPA/Xinhua; Science Advances; IEEE Spectrum; ClinicalTrials.gov (checked on 4 Nov 2026); FDA (510(k) K242618); Vivani Medical, 2 Jul 2026; Eduardo Fernández (UMH), 22 Sep 2026.
Why caution pays: the Second Sight case
There is a concrete reason not to read these promises as a guaranteed purchase, and it has names. Second Sight gave partial sight back to more than 350 people with its Argus II implant. In 2019 it announced it would stop making the device, in 2020 it laid off most of its staff and announced it would wind down operations, and in 2022 it merged with another company and vanished as such, leaving those patients with an implant in the eye, no spare parts and nobody to repair it (IEEE Spectrum documented it). With one qualification that has to be made: what died was the retinal implant. The house’s cortical programme, Orion, lives on at Cortigent, a Vivani Medical subsidiary awaiting a merger with the listed company ClearOne (agreed in July 2026): it implanted six blind people between January 2018 and January 2019, completed its feasibility study in 2025, according to the company, and presented six-year results at the NANS conference in January 2026, with improvements on visual function tests with the system on versus off and fewer than 4% of electrodes losing function. Terry Byland, the company’s public face for years, is not one of them: he carries the retinal implant —an Argus I in his right eye in 2004 and an Argus II in his left eleven years later, the first patient in the world with two— and he is one of those left without support. This is how he summed up his position to IEEE Spectrum in 2022: “As long as nothing goes wrong, I’m fine. But if something does go wrong with it, well, I’m screwed. Because there’s no way of getting it fixed.”
Fantastic technology and a lousy company.
Ross Doerr, Second Sight patient
The useful question about an implant is not only whether it works, but whether it will still be working when the company changes its plans.
What the man who leads it says
Eduardo Fernández Jover, who leads the Elche team, asked when the study came out for progress little by little and no false expectations, according to the press release from the Universidad Miguel Hernández.
On the phone, on 22 September 2026, he explains why he insists on it: “When someone has a problem, an accident for instance, they go to the internet, ask artificial intelligence and find several groups. That can raise false hopes that don’t match reality. This is pure research, and although it is moving forward, you never know.” That is why his team points blind people to the ONCE, Spain’s national organisation for the blind: “The ONCE already has tools. This may work and it may not.”
And he draws the line: “Recovering sight the way people who are not blind understand it is, for now, not viable.” What they have achieved is something else: “The progress we see comes more from the patient’s interpretation in their brain than from vision as such.”
Nor does it work for everyone. “Most blindness is in the eye, from something genetic, for example. But there are several bottlenecks to solve, and each case of blindness can have different causes.” A cortical implant only makes sense when the damage lies further back, which is what Chart 4 shows. The work goes on: counting the patient they are working with now, his team has implanted five people.
The wider frame comes from another Spaniard, the neurobiologist Rafael Yuste, who drove the idea of “neurorights” and is one of the voices behind the León Declaration on European neurotechnology, issued in 2023 during Spain’s presidency of the EU Council: “The brain has to be protected as the sanctuary of our mind, because that is where human identity is generated.” He is not talking about this case in particular, but about the ground it is played on.
What to watch to know whether this is really moving
Four signals, above the noise: that Neuralink registers Blindsight on ClinicalTrials.gov and publishes it in a peer-reviewed journal; that the Elche team goes from individual cases to a larger clinical trial; that China publishes results from its patients beyond the regulatory announcement; and, above all, the real resolution (making out shapes is not the same as seeing). Until those boxes are ticked, every new deadline is a date, not a fact.
Where each implant acts: eye → optic nerve → cortex
The chain of vision, and the point where each implant steps in.
Science Advances (UMH Elche); IEEE Spectrum and Vivani Medical (Argus II). Schematic illustration.
Next time Musk promises to restore sight imminently, the useful question is not whether he will. It is who is already testing it (in Elche, with implants that are taken out after six months) and what the man who leads the work asks for: no false expectations. And, if the day comes, that the implant stays switched on long after the spotlights go out.
Correction · 4 November 2026
Corrected facts
Correction. This piece was published on 11 August 2026. On 4 November 2026 its facts were checked against their sources and the following was changed:
- Headline: it said a Spanish team “already has” restored sight. In Elche they test implants that are taken out after six months and make people perceive points of light. It now says they are testing it.
- Musk gave the six-to-twelve-month deadline at Y Combinator in June 2025, not in August 2026.
- Removed: the promise of 3 August 2026 on Ashlee Vance’s podcast and the press headline “in the coming months”. The podcast’s archive has no episode with Musk, and no source was found.
- At Y Combinator, Musk did not compare the resolution to “the graphics of the first Nintendo”. He said it would be quite low at first.
- Science Advances (2025): “patterns and movement” comes from the university’s press release, not the paper, and is now attributed to it. In the paper, adjusting stimulation to the brain’s response is a future step; the piece presented it as done.
- The line from Fernández about moving forward little by little and not creating false expectations was not a direct quote: the UMH press release gives it in reported speech. It now reads that way.
- Brain Communications (2026): added that the improvement was in his own sight, that it declined after the implant was taken out, and that he remains within the blindness range.
- Chart 3: the clinical trial counter said 4 and there were 5 rows. Neuralink also has a trial in the UK. Synchron opened three more trials between March and May 2026, one in Canada.
- Orion: March 2025 was an estimated date in the trial registry. The company says it completed its feasibility study in 2025.
- Cortigent is still a Vivani subsidiary, with a merger with ClearOne agreed in July 2026 and not yet closed.
- BrainGate is no longer the largest trial: Neuralink had 21 participants in January 2026. It is still the longest.
- Terry Byland: his quote is from February 2022, and it now uses his exact words in IEEE Spectrum. The Argus II sits in the eye, not inside the head.
- Second Sight: in 2020 it laid off most of its staff and announced it would wind down operations. The piece said it stopped supporting the device.
- Photoreceptors: it said between 120 and 130 million. The count by Curcio et al. (1990) gives an average of 92 million rods and 4.6 million cones per retina.
- Chart 4: the Argus II was approved for a rare form of retinal blindness. It said it worked whenever the retina was damaged.
- Removed: that plenty of neuroscientists are comfortable with the goal of Blindsight. It had no source.
- What is known about Blindsight also comes from Neuralink’s website, not only from Musk’s X account.
- Sources: the 21 participants come from Neuralink’s website, not from Reuters. Neuralink announced the FDA designation; the FDA does not publish these designations.
- Chart 1: Musk spoke at Y Combinator on 16 June 2025. The 19th is the day the video was published.
- The Blindsight zeros (patients, peer-reviewed studies and registrations) were checked again on 4 November 2026.
- The counter in Chart 1 counted the days up to the reader’s own date and said Blindsight was “today” still on 0 patients. It now counts up to 4 November 2026, the day of the check.
Sources checked on 4 November 2026: Y Combinator video of 19 June 2025; the Core Memory podcast archive; ClinicalTrials.gov; PubMed; neuralink.com/updates; full text of Science Advances (2025) and Brain Communications (2026) on PMC; UMH press release of 6 November 2025; IEEE Spectrum, 15 February 2022; Vivani Medical press release of 2 July 2026; ClearOne’s SEC filings; Curcio et al., J Comp Neurol, 1990.
Sources
- UMH / Dr Balmis Hospital (Alicante, Spain) — Science Advances, Nov 2025. DOI 10.1126/sciadv.adv8846. Primary.
- Neuralink, “Two Years of Telepathy”, 28 Jan 2026 (21 participants) — neuralink.com. Primary / company statement.
- Neuralink — announcement of the FDA Breakthrough Device designation for Blindsight, 17 Sep 2024 (the FDA does not publish these designations) — neuralink.com. Primary / company statement.
- FDA — 510(k) K242618, Precision Neuroscience’s “Layer 7-T”, cleared on 30 March 2025: temporary use (under 30 days), assessed with bench testing and an animal study, no clinical data. Primary.
- Rubin DB et al., “Interim Safety Profile From the Feasibility Study of the BrainGate Neural Interface System”, Neurology 100(11):e1177-e1192, 14 March 2023 — DOI 10.1212/WNL.0000000000201707. 14 adults implanted between 2004 and 2021 in BrainGate (NCT00912041), the longest brain implant trial. Primary.
- NMPA (China) / Xinhua — approval of NEO, 13 March 2026. Primary.
- IEEE Spectrum, “Their Bionic Eyes Are Now Obsolete and Unsupported” (2022) — Second Sight / Argus II, more than 350 patients implanted; quotes from Terry Byland and Ross Doerr (“fantastic technology and a lousy company”). Solid secondary.
- Public statements by Elon Musk on X, at Y Combinator (video of 19 Jun 2025) and at the Qatar Economic Forum (Sep 2024–Jan 2026) — a primary but self-interested source: no ClinicalTrials.gov registration and no peer-reviewed publication behind them (see “What is actually true at Neuralink”).
- Rafael Yuste — published interviews; León Declaration (2023, Spanish presidency of the EU, 30+ countries). Secondary.
- Universidad Miguel Hernández, press release of 6 Nov 2025 on the Science Advances study — comunicacion.umh.es. Primary / press release.
- Alfaro A, Soo L, Waclawczyk D, Morollón R, Grani F, Fernández E, “The unexpected sight: improvement of visual function following intracortical microstimulation of the human occipital cortex”, Brain Communications 8(1), 3 Feb 2026 — DOI 10.1093/braincomms/fcaf504. One participant; implant removed after six months, improvement still present eighteen months later. Primary.
- Eduardo Fernández Jover, phone conversation with the author, 22 Sep 2026: shapes rather than points, the dialogue with the brain, the ONCE, the limits of the technique and the five people implanted, a figure he later confirmed by email (it includes the current patient). Primary.
- Curcio CA et al., “Human photoreceptor topography”, J Comp Neurol 292(4):497-523, 1990 — DOI 10.1002/cne.902920402. Photoreceptors per retina. Primary.
- Vivani Medical, press release of 2 Jul 2026 — merger agreement between Cortigent and ClearOne; Argus II, approved for a rare form of blindness — investors.vivani.com. ClearOne’s SEC filings up to 15 Oct 2026: the merger is not recorded as closed. Primary.
- ClinicalTrials.gov, checked on 4 Nov 2026: Neuralink’s seven studies (none on vision), Synchron’s trials, the UMH study and the Orion study. Primary.