Human Neurons Are Now Renting Compute Time. Nobody Has Written the Rules.

What actually exists, corrected against the company's own claims

An Australian company, Cortical Labs, sells access to computers that run on living human brain cells. Its CL1 device grows, per independent trade-press reporting corroborating the company's own technical materials, roughly 200,000 lab-cultured neurons derived from adult donors' blood or skin cells, not embryos, directly onto a silicon microelectrode array, then keeps them alive with a life-support system and interfaces with them through software the company itself calls "biOS," which its own site describes as part of "the first biological computer enabling medical and research labs to test how real neurons process information." This has been a real, purchasable product since 2025, at roughly 35,000 US dollars a unit, and remote access to living neurons has been rentable through the company's Cortical Cloud service for even longer. What is new in 2026 is scale: Cortical Labs has aggregated its own hardware into what it calls biological data centres, with a 120-unit facility running in Melbourne, confirmed operating by March 2026, and a 20-unit prototype rack unveiled at Singapore's National University of Singapore Medicine on 6 August 2026, with a stated ambition to reach 1,000 units there, subject to technical validation and regulatory approvals that have not yet been granted.

It is worth being precise about what is genuinely novel here versus what recent press coverage has overstated. This is not the first neuron-based computer, that dates to a 2022 paper in the journal Neuron from the same company; it is not the first commercial sale of the hardware, that was 2025; what is new is running it at data-centre scale as aggregated commercial infrastructure. One widely reported recent demonstration, of the same system playing a simplified version of a video game, has not been published in a peer-reviewed venue, unlike the original 2022 result, and should be read as a company demo, not a scientific finding, until that changes.

Why this exists at all

The underlying motivation is the same one behind AI's broader energy pressure that this site tracks in its cloud-emissions coverage. The human brain runs on roughly 20 watts. The IEA's most recent figures, from 16 April 2026, show global data-centre electricity demand grew 17 percent in 2025 against 3 percent growth in overall electricity demand, with AI-specific data-centre consumption on track to triple by 2030 against a doubling for data centres overall, and the five largest tech companies' combined capital expenditure exceeded 400 billion US dollars in 2025 alone. That gap between silicon's power appetite and the brain's is precisely what is pulling serious money into every alternative computing paradigm right now, biological and otherwise. It is worth being honest that Cortical Labs' own efficiency and capability claims for CL1 remain, in the company's own framing and in every independent account found for this piece, speculative rather than demonstrated at any meaningful scale.

The actual governance story: two real ethics frameworks, and neither covers this

A real academic field already exists for the ethics of exactly this kind of biological system, Organoid Intelligence, named at a February 2022 Johns Hopkins-led workshop whose findings were published in two companion 2023 papers, one of which, in Frontiers in Science, is the piece most specifically known as the Baltimore Declaration. Together they call for a rigorous process of consent and re-consent from tissue donors, explicit vigilance for any sign a cultured system is developing consciousness, and ethicists embedded directly in research teams rather than bolted on as external reviewers. The UK's Nuffield Council on Bioethics published its own independent assessment of neural organoid research in March 2024, calling for future-looking and proportionate regulation and honest public communication about what these systems can and cannot do.

Here is the gap. All of this, the most serious governance work this field has produced, is written for research use. None of it addresses a stem-cell-derived neuron culture deployed as commercial computing infrastructure, rented out by the hour to paying customers, which is exactly what Cortical Cloud already does today. Some of the field's own leading researchers have said so explicitly: a November 2025 Science commentary by Stanford's Sergiu Pasca alongside bioethicists Insoo Hyun and Paola Arlotta names the gap directly, with Hyun asking pointedly whether stem-cell donors should be told how they would feel having their donated cells plugged into a biocomputer, and calling for a new international oversight body because existing frameworks do not reach commercial use. As of a follow-up report the same group published in September 2026, that body still does not exist, and the researchers say plainly that no globally standardised ethical or regulatory framework covers this yet.

Australia's own stem-cell oversight regime, built around the Prohibition of Human Cloning Act and the Research Involving Human Embryos Act, both administered through NHMRC-licensed Human Research Ethics Committees, is constructed around embryonic tissue and research or therapeutic use. Because CL1 uses adult-donor-derived induced neurons for commercial infrastructure rather than embryonic tissue for research, it plausibly sits outside that regime's clearest triggers entirely, which would make ethics review a matter of institutional choice rather than legal requirement once a device leaves the lab and goes into a commercial rack. Singapore's own regulators appear to recognise the gap in practice: the joint statement behind the NUS Medicine facility explicitly describes jointly establishing governance, biosafety, and compliance frameworks as a live, unfinished task, and the larger 1,000-unit build-out is explicitly contingent on approvals that do not yet exist.

The field's own scientists are already worried about this

This is not outside criticism looking for a reason to be alarmed. STAT News reported in November 2025 that researchers at Stanford, Cambridge, and Cold Spring Harbor Laboratory have publicly warned that overstated claims about these systems, particularly the word "sentience," which was used in the 2022 research paper's own title and drew a formal published rebuttal from thirty other researchers, risk triggering a regulatory backlash severe enough to damage legitimate medical organoid research that has nothing to do with computing. Cortical Labs' own chief scientist is on record making a related point from inside the company: if researchers cannot agree on shared language for what these systems are, the ethics conversation cannot get anywhere. A field whose own most prominent commercial figure says the vocabulary is not settled yet is not a field that is ready to be treated as routine infrastructure.

What responsible disclosure and governance would look like here, today

Nobody is required to do any of the following yet, which is exactly the point.

• Publish stem-cell donor consent terms specifically for commercial, not just research, downstream use, since the two existing ethics frameworks were both written assuming a research context.
• Commission an independent welfare and sentience assessment on a fixed schedule as culture scale and duration increase, CL1 units currently remain viable roughly six months, rather than waiting for a scandal to force the question.
• Separate verified, peer-reviewed capability claims from company demonstrations explicitly in all marketing, so customers and regulators are not reasoning from promotional footage.
• Treat Singapore's still-being-established governance moment as the template, not the exception: any jurisdiction approached about hosting this technology should ask, before infrastructure is built, exactly what Singapore's regulators are still working out after the fact.

The honest bottom line

Nothing here suggests Cortical Labs is acting in bad faith; if anything, its own chief scientist's candour about the field's unsettled vocabulary is a point in its favour. But a technology that uses living human-derived tissue as rentable commercial infrastructure has, right now, two serious ethics frameworks that do not apply to its actual commercial use, an Australian regulatory regime that plausibly does not reach it, and a Singaporean regulator that says openly it has not finished writing the rules for the very facility already partly built. That is not a reason to stop the research. It is exactly the moment, before the technology reaches the scale its backers are aiming for, when the governance work is cheapest to do and most likely to be listened to.

Sources: Cortical Labs, CL1 product page; NUS Medicine Singapore prototype announcement; Nuffield Council on Bioethics; STAT News; IEA data-centre electricity release.