Implantable BCI Ethics
Ethical framework for implantable human brain–computer interface research, proposed by Miller, Giampiccolo, Akram & Schalk in a Nature Neuroscience Comment (7 Sep 2026), written as human implantations expand and “ethical clarity must keep pace with technical ambition.” The Comment draws on the authors’ earlier editorial — Miller & Abosch, “A Moment of Reckoning for Implanted Brain–Computer Interface Studies” (Neurosurgery, 2025) — which carries the detailed argument and is openly available via PMC.1
The four proposed commitments
From the Comment’s abstract:1
- Distinguish research participation from patient care.
- Ensure long-term support for study participants.
- Ensure participants do not bear disproportionate risks for benefits realized mainly by others.
- Ground research in meaningful clinical purpose.
Research subject or patient? (the core distinction)
Conflating research subjects with patients is “more than a semantic concern.” If a study is not designed to offer direct or even plausible therapeutic benefit to the individual, that person is a research participant, not a patient — yet across the literature, media, and even consent processes the distinction is frequently obscured. The risk is highest for vulnerable populations (ALS, high cervical spinal cord injury). The authors warn against therapeutic misconception: participants may describe the experience as positive or uplifting, but structured engagement can feel rewarding against otherwise isolating circumstances — psychological uplift is not functional gain and should not be marketed as participant satisfaction. If there is no therapeutic link to the participant’s condition, researchers should consider using healthier volunteers instead, which minimizes ethical complexity. The argument that severely disabled participants “have less to lose” is flawed — especially for penetrating electrodes in eloquent cortex, where limited residual function may be at stake.2
What justifies an implanted BCI study?
For implantation in humans to be ethical, a study should offer at least one of:2
- Direct therapeutic benefit to participants;
- Generalizable scientific knowledge relevant to future therapies;
- A credible pathway toward scalable, sustainable device translation.
If none are present, the primary outcome may be limited to media attention and academic prestige. Demonstrations of drone control, avatar manipulation, or social-media posting from an implant may be headline-grabbing but do not correlate with principles that guide device development or clinical translation.
Usability and open access: beyond the laboratory
All in-human implanted BCI systems should support some form of independent use in the participant’s home, for their own purposes, without researchers present — yet most systems operate only in research contexts under technician supervision. Instead of bespoke demonstrations, translational software should generate surrogates for keyboard, mouse, and eye-tracker function and interoperate with validated commercial assistive applications (Tobii, Wego, AbleNet, built-in OS features) — which are already familiar to rehabilitation infrastructure and sustained by a viable market. Studies must also explicitly disclose usability limitations up front: likely implant longevity, technical-support duration, and hardware/software support limits. Most penetrating electrode systems in humans lose signal quality within months to years yet remain implanted in the hope of long-term benefit; real impact requires manufacturable, maintainable device ecosystems with standardized clinical workflows and provider training — not one-off platforms.2
Risk, morbidity, and explantation
Every implanted device carries risk — infection, hemorrhage, neurological deficit, long-term morbidity — which must be communicated clearly and pragmatically by a noninterested party, ideally with illustrations of chronically exposed hardware, incisions, and potential cosmetic morbidity. Consent should cover the risks of explantation, which can exceed those of implantation (arachnoidal encapsulation, vascular ingrowth, dural or calvarial overgrowth, wiring incorporation into galea and bone). Some protocols mandate routine explantation after trial completion — a practice rarely seen in other implantable therapies — requiring reoperation that can escalate from burr holes to full craniotomy. 2
Durability and device support: who is responsible?
Many BCI trials are backed by startups of uncertain longevity. If a company fails, participants may be left with orphaned devices — the implanted-device analogue of the Argus II abandonment (Second Sight’s 2019–2020 wind-down left >350 implantees unsupported). The authors propose three contingency measures:2
- Connector interoperability mandates — hardware compatible with current and future alternatives;
- Escrow funds or written hospital pledges — secured at implantation, covering device support or removal;
- Predefined sustainability plans — approved by oversight boards for startup-backed research implants.
Trial approvals should include a transition plan if devices are orphaned, and the plan should be part of the consent process. If industry partners cannot guarantee long-term support, investigators and ethics boards must be transparent about this during consent. 2
Scientific justification: toward generalizable knowledge
When studies involve neurosurgical implantation in individuals with severe disabilities, the ethical threshold for justification is substantially higher. Absent direct benefit, the insights gained must be broadly applicable: which brain regions generate reliable control signals, how feedback shapes cortical activity, which electrode scales match each biomarker, biomarker longevity, pathophysiology constraints, and neural correlates of intention. Incremental replications, superficial outputs, or closed systems that withhold data fail this test: “We are either constructing the conceptual pyramid of brain-computer interfacing — laying bricks on which future therapies can be built — or we are distributing rubble.” 2
Summary of recommendations (editorial Table 1)
| Area of focus | Recommendation |
|---|---|
| Participant vs patient | Clearly distinguish research participants from clinical patients; avoid therapeutic language unless a clinical benefit is intended |
| Informed consent | Explicit discussion of risks (craniotomy, explantation), expected usability, long-term sustainability; visual depictions of hardware/scars |
| Device usability and integration | Function outside laboratory settings; interoperate with existing assistive technologies used by the target population |
| Scientific contribution | Design for generalizable knowledge (signal viability, feedback plasticity, cortical integration) |
| Risk–benefit justification | Require therapeutic intent or significant expected scientific insight; avoid implantations that only serve as demonstrations |
| Device sustainability | Contingency plans for device support and interoperability, especially for non-commercial devices |
| Ethical framing | Avoid over-promising; ensure media and recruitment materials do not blur research and therapy |
Wiki context
- The Argus II episode (retinal-prosthesis, cortigent) is the canonical real-world failure of the long-term-support principle; the IEEE Spectrum investigation documented patients left unable to obtain MRI scans, crowd-sourcing spare parts, and unsupported implants. 3
- Co-author Harith Akram is Chief Investigator of Neuralink’s GB-PRIME study — disclosed as a competing interest in the Comment (see neuralink).
- The U.S. GAO’s late-2024 BCI report addressed data ownership, insurance, and long-term support, but emphasized second-order issues (cybersecurity, interoperability) over physical morbidity and ethical enrollment — a direct prompt for both papers.
- The Comment sits in the Belmont Report lineage of research-subject protections, applied to a field where participants are often also patients and companies may fail mid-study.
- Visual-prosthesis ethics were systematically reviewed by van Velthoven, …, Chen X, Roelfsema PR (J Neural Eng 2022) — see xing-chen; the Miller et al. framework generalizes that concern to BCI studies as a class.
Open questions
- Adoption gap: the interoperability, escrow, and sustainability-plan measures are proposals, not current regulatory requirements — whether IRBs, funders, or regulators will mandate them is unresolved.
- Can “meaningful clinical purpose” be operationalized as an enrollment criterion without blocking early-stage science?
- How should risk–benefit accounting change when the beneficiary is mainly future patients or a company rather than the participant?
Related pages
- brain-computer-interface — category hub
- brain-interface — cortical/cerebral subcategory: the main site of implanted trials
- retinal-prosthesis — the Argus II abandonment precedent for long-term-support failure
- cortigent — device-support failure case (Argus II legacy)
- neuralink — GB-PRIME (disclosed interest of co-author H. Akram)
References
- Miller KJ, Giampiccolo D, Akram H, Schalk G. Ethical considerations for implantable human brain–computer interfaces. Nat Neurosci (2026). DOI: 10.1038/s41593-026-02447-1. 1
- Miller KJ, Abosch A. A Moment of Reckoning for Implanted Brain-Computer Interface Studies. Neurosurgery 2025;97(2):277–280. DOI: 10.1227/neu.0000000000003585. PMC12709552. 2
- Related literature cited therein: GAO, Brain–Computer Interfaces: Applications, Challenges, and Policy Options (2024); Belmont Report (1979); Vansteensel et al., NEJM 391:619–626 (2024); Goering et al., Hastings Center Rep 54:24–33 (2024); Schönweitz et al., Brain Stimul 17:1145–1154 (2024).