Picture the scene: a researcher in Leipzig or Seoul opens a Jupyter notebook, imports a Python library, and within seconds is sending electrical pulses to living human neurons sitting in a bioreactor in Vevey, Switzerland. No shipping, no lab coat required. For about the price of a mid-tier GPU instance, you can now rent a piece of a living human brain; not a metaphor, not a neural network, but actual lab-grown human neurons, alive in a bioreactor, that you talk to over the internet using Python. That is not a thought experiment from a bioethics seminar. That is the product catalog of FinalSpark, a Swiss startup whose Neuroplatform has been commercially live since May 2024. FinalSpark launched the world's first commercially accessible biocomputing research platform on May 15, 2024, and it is operational right now in Vevey, Switzerland. Academic customers can get access to the platform, featuring four shared organoids, for $500 per user per month, or even free for selected projects, and for the fee FinalSpark says users get to conduct biocomputing research on a 24/7 fully managed remote neuroplatform. Meanwhile, on the other side of the planet, Melbourne-based Cortical Labs has gone further still. The CL1, priced at $35,000, integrates living neurons with silicon for real-time computation. Designed to keep neurons alive for up to six months with its internal life support system, the CL1 needs minimal inputs and a fraction of the energy other technologies use, enabling longer research timelines. The company has even announced Doom running on a CL1 built with the Cortical Cloud. Living brain tissue is now a line item on a purchase order.
The road to this moment runs through a petri dish and a 1970s arcade game. In 2022, a cluster of roughly 800,000 human neurons grown in a petri dish learned to play Pong; not metaphorically, literally; and within five minutes of receiving feedback signals, the dish brain began improving its paddle control. Within weeks it could sustain rallies, and the experiment, dubbed DishBrain, forced neuroscientists to rethink what learning is, what neurons fundamentally want to do, and whether biology might eventually outperform silicon at its own game. That was the proof of concept. By early 2026, Cortical Labs had escalated: brain organoids were moving, aiming, and shooting in a feedback loop driven entirely by biological neural activity in experiments that began circulating publicly in early 2026. The underlying pitch driving all of this is energy. Training the AI system that mastered the board game Go required roughly 4 times 10 to the power of 10 joules of energy; about a decade of human metabolic energy; whereas the human brain consumes only 20 watts, roughly the power of a dim light bulb. FinalSpark has even built a dopamine-reward loop into its platform: when an organoid produces a desired output, the system pulses UV light to uncage dopamine into the tissue, the chemical equivalent of 'good job'; it is, functionally, reinforcement learning with neurotransmitters. The efficiency argument is real, the commercial momentum is undeniable, and the ethical scaffolding is almost entirely absent.
That last point is the subject of a piece published in Nature on July 27, 2026, carrying the DOI 10.1038/d41586-026-02316-8 and authored by a team that includes bioethicists and engineers from Harvard Medical School, the University of Michigan, Indiana University, and Hebrew University of Jerusalem. The core finding is strikingly simple and strikingly damning. Discussions have raged over the morality of using lab-grown brain tissue and the potential for consciousness, yet one key issue has been absent from debates so far: the lack of explicit consent for the use of human neural tissue in biocomputing. People who give tissue samples for biomedical research might have no idea that their cells are being used for biocomputers. That consent gap is not a technicality. Donors who handed over stem cells for, say, Parkinson's research almost certainly did not authorize those cells to be wired into a commercial computing substrate, kept alive for months, stimulated with electrical patterns, rewarded with dopamine pulses, and rented out by the API call to anyone with a credit card. The Nature authors propose a concrete fix: new review boards could be established at institutes undertaking biocomputing, or reconfigured from existing stem-cell-research oversight committees supplemented by neuroscience and computer-engineering scholars, and these committees should instruct researchers to follow one of three pathways in strict order: establish new cell lines, obtain fresh consent from existing donors, or undergo independent review of research plans. No such boards exist today at any of the major commercial platforms.
The consent problem is urgent and tractable. The consciousness problem is urgent and not. The field's standard reassurance, delivered in papers and press releases with remarkable uniformity, is that current organoids are nowhere near conscious and the question is therefore premature. While complex network behavior is beginning to emerge even without much external stimulation, experts generally agree that current organoids are not conscious, nor close to it. But a growing cohort of researchers is now arguing that this reassurance is doing structural work it cannot honestly support. A 2025 paper in Patterns, cited in the 2026 Nature literature, spelled out the dynamic with precision. Perhaps fearing that consciousness-related concerns and related restrictions might stifle scientific progress, neuroscientists and philosophers have mobilized in many post-2020 publications to defend organoid research, claiming a neuroscientific consensus that dismisses anything but a far-distant potential for consciousness as biologically implausible; in doing so, they have positioned any current reassessment of use criteria, or calls for additional safeguards, as unnecessary. A March 2026 paper in Scientific Reports examined public moral intuitions and found a more complicated picture: the public's tendency to attribute consciousness to biocomputers is positively correlated with attribution of other cognitive traits and increased ethical concerns, but also with perceived benefits and overall support for research. In other words, people who think these systems might feel something are also the people most excited about them, which is not the kind of moral clarity the field is hoping for.
The regulatory landscape is almost as sparse as the ethical one. Guidelines published by the International Society for Stem Cell Research focus on the current state of organoids as being devoid of human consciousness, but without much speculation for the future; today's human brain organoids may be far from having such capabilities, nevertheless, dismissing concerns about moral status provides unstable ground on which to build a global human brain organoid policy. The one jurisdiction that moved first is China: on April 29, 2025, China's National Science and Technology Ethics Committee issued the Human Organoid Research Ethical Guidelines, establishing the world's first comprehensive governance framework especially focusing on brain organoids, embryo models, and chimeric research. The United States and European Union have produced no equivalent. Ethical and regulatory frameworks for human-neuron computing are still developing. Meanwhile, FinalSpark's own roadmap calls for commercial pilot programs from 2026 to 2028, partnering with pharmaceutical companies for drug discovery applications. The gap between what the technology is doing and what the rules say it can do is not a gap that closes itself; it widens under commercial pressure, and right now the pressure is all pointing one way.
The 'not conscious, not close' reassurance has been the field's load-bearing ethical beam since 2020; if the science shifts even slightly, and the commercial scaling continues, there will be nothing underneath it.