In February 2026, a biocomputing company demonstrated 200,000 living human neurons playing Doom.[1] The demonstration used a commercial product advertised at the time with a Python API and a price around $35,000.[2] An independent developer built the Doom integration in about a week; the operator's original Pong work had taken more than 18 months on its original hardware and software.[1] Separately, a 2024 paper described a Swiss company's remote research platform with sixteen organoids and Python API access, then offered free for research.[3]
The use of donor-derived tissue raises questions about provenance and permitted uses. The public materials reviewed for this April essay did not let us trace particular workloads to particular donor permissions. That is a limit of this review, not evidence that consent was absent or breached. Current pricing, access terms and institutional arrangements require a separate check.
The evidence is peer-reviewed, extended by an independent group, and commercially deployed.
In 2022, Kagan et al. published the DishBrain system in Neuron, showing that human iPSC-derived neural cultures exhibit adaptive, goal-directed learning in a Pong environment within five minutes of play (Kagan et al., Neuron 2022).[4] The cultures learned through stimulation and feedback, in a pattern consistent with biological learning.
A year later, an independent group extended the work to a different task. Cai et al. at Indiana University published Brainoware in Nature Electronics, demonstrating brain organoid reservoir computing for speech recognition and nonlinear equation prediction (Cai et al., Nature Electronics 2023).[5] Different lab, different platform, same conclusion: biological neural networks perform real computational tasks.
By 2024, a company had turned the research into remotely accessible infrastructure. Its Neuroplatform, described in Frontiers in Artificial Intelligence, runs sixteen organoids around the clock with Python API access; over three years it used more than 1,000 organoids and collected more than 18 terabytes of data, and in 2024 access was free for research (Frontiers in AI, 2024).[3] Researchers elsewhere can run experiments on living human neurons remotely.
Then there is the finding that makes the governance question urgent. Gabriel et al. published in Cell Stem Cell in 2021 that iPSC-derived brain organoids spontaneously develop bilaterally symmetric optic vesicles containing light-responsive photoreceptors, with axonal projections connecting to forebrain regions (Gabriel et al., Cell Stem Cell 2021).[6] Across 16 independent batches from four iPSC donors, the researchers generated 314 brain organoids, 72% of which formed optic cups, according to the publisher's release on the paper.[6] These substrates developed complex sensory structures that were not present at initialization and were not directed by the operators. The governance question does not depend on whether biological computing scales to data-center size. It depends on whether third-party workloads are being processed on donor-derived tissue.
The Governance Gap
Who governs what these neurons are permitted to compute? Our March 2026 public-document review did not establish the complete consent, ethics-review, workload and decommissioning arrangements of the platforms discussed. Public product documentation is not the whole institutional record. This essay proposes a way to organize evidence; it does not establish that the field has no governance.
The HeLa history illustrates why provenance and consent deserve scrutiny. Cells taken from Henrietta Lacks in 1951 without informed consent were commercially distributed for decades before her family was notified.[7] That history does not establish a consent failure in the contemporary platforms discussed here; their institutional arrangements require their own assessment.
Biological research already has ethical and institutional governance. The Baltimore Declaration calls for ongoing, multistakeholder discussion of organoid intelligence, including donor interests and possible consciousness.[8] A cryptographic record could support parts of that work. It cannot determine whether consent is adequate, replace ethics review or turn a recorded assertion into a biological fact.
Why This Is Different
Biological computing connects digital workloads with living, donor-derived material. That makes provenance, consent scope and oversight important questions for the responsible institution. A change in a culture or its use may warrant review against the applicable permissions and protocol. It does not automatically establish a consent violation, and the organoid findings cited here do not establish consciousness in a commercial computing platform.
The substrate can change during an experiment. Our proposed term, consent drift, describes a possible mismatch between a recorded permitted use and a later use or capability. Whether such a mismatch exists requires biological, ethical and legal judgment. A record system can preserve the relevant decisions; it cannot make those judgments by comparing hashes.
End-of-life handling also requires a protocol appropriate to the material and the institution. A digital system could retain who authorized decommissioning, what procedure was reported and what supporting evidence was collected. A signature authenticates the signed record. It cannot establish that no viable cells remain or decide what disposal method is required.
Our proposed architecture uses signed records at selected biological-digital interfaces to link provenance, permitted uses and recorded decisions. This is one design option, not a demonstrated requirement for every adequate governance system. Record durability, independent retention and cryptographic migration should be assessed over the intended retention period.
Long-lived records also need an explicit threat model. The March 2026 quantum resource estimate concerns hypothetical hardware and secp256k1 keys, not a demonstrated attack on a deployed biological system or a deadline for one.[9] It motivates evaluating migration and preservation strategies; it does not prove that a particular substrate requires a particular signature scheme today.
We have been working on this problem at Attested Intelligence. The Attested Governance Artifacts architecture could extend to biological substrates, with bio-compute-specific additions for tissue provenance, consent drift monitoring, and verified end-of-life. That design is described in the paper “Cryptographic Governance for Biological Computing” (WP-AIH-2026-001, v8.0, available on request): schemas, a conformance checklist, and an eleven-phase lifecycle from donor consent through tissue deprovisioning. A public repository for the schemas is planned; it is not published yet, and none of the bio-compute additions ships in the AGA packages.
For this proposal, the next step is domain review: can a record design help a qualified institution reconstruct permitted use and oversight without implying that cryptography validates consent or biological outcomes? No biological pilot, deployed integration or validated consent-monitoring capability is claimed.
The published experiments make biological computing a serious research subject. They do not, by themselves, validate AGA's proposed extensions or establish an absence of existing oversight.
The practical question is what evidence a responsible reviewer would need, who can supply it, and which parts a digital record can honestly support.
References
- Demonstration video of living human brain cells playing DOOM on a commercial biocomputer. YouTube, February 25, 2026. Press coverage: Tom's Hardware (March 1, 2026), PC Gamer (March 9, 2026), and Popular Science, “Computer run on human brain cells learned to play ‘Doom’” (March 2, 2026), which gives the one-week and 18-month figures.
- Product pricing (~$35,000): the operator's product page; Top Gear (March 2026).
- “Open and remotely accessible Neuroplatform for research in wetware computing.” Frontiers in AI 7:1376042, 2024. DOI:10.3389/frai.2024.1376042.
- Kagan BJ et al. “In vitro neurons learn and exhibit sentience when embodied in a simulated game-world.” Neuron 110:3952-69, 2022. DOI:10.1016/j.neuron.2022.09.001.
- Cai H et al. “Brain organoid reservoir computing for artificial intelligence.” Nature Electronics 6:1032-39, 2023. DOI:10.1038/s41928-023-01069-w.
- Gabriel E et al. “Human brain organoids assemble functionally integrated bilateral optic vesicles.” Cell Stem Cell 28(10):1740-57, 2021. DOI:10.1016/j.stem.2021.07.010. The batch, donor and optic-cup figures are from the Cell Press release, “Brain organoids develop optic cups that respond to light,” EurekAlert!, August 17, 2021.
- Skloot R. The Immortal Life of Henrietta Lacks. Crown, 2010.
- Hartung T et al. “The Baltimore Declaration toward the exploration of organoid intelligence.” Frontiers in Science 1:1068159, 2023. DOI:10.3389/fsci.2023.1068159.
- “Securing Elliptic Curve Cryptocurrencies against Quantum Vulnerabilities: Resource Estimates and Mitigations.” arXiv:2603.28846, March 30, 2026. A conditional resource estimate for hypothetical quantum hardware, not an observed break; revised April 15, 2026.
See the working implementation on npm.
AGA is a reference implementation of a published format for verifiable decision records: hash-bound policies, signed Decision Receipts, and Evidence Bundles that verify offline. The implementation is on npm. The evaluation path walks through it in working code.
RecordThe paper WP-AIH-2026-001 (v8.0): published April 2026. The original document is preserved. Its proposals and descriptions may differ from the published implementation. For current capabilities and limits, see Trust and scope.
The full paper, “Cryptographic Governance for Biological Computing,” is available on request. Related: Two Threats No Dashboard Can See | What signed records add to a deployment review | Who Controls the Model at Runtime?