Nothing on sale today lets a healthy person think faster, see further or move a third arm. RAND Europe wrote the rulebook for the day something does.
Governing Augmentative Brain-Computer Interfaces was released on 2 November 2026 by RAND Europe and the Brussels think tank Centre for Future Generations. The report runs 55 pages and repeats one finding without hedging: no commercially available BCI has demonstrated robust augmentation beyond normal human capability under real-world conditions.
Pole one: test the use case, not the device
The case for acting now rests on signals. Nine structured interviews with researchers and company staff in Germany, the US, the UK, Sweden and Spain fed a Brussels workshop in June that tested seven scenarios, from a neurosurgeon recovering from a stroke to a classroom of students in EEG headbands.
Those experts expect the first real enhancement in control of external devices. Non-invasive speech decoding comes next. New senses stay a medium-to-long-term prospect for implants. Cognitive and emotional enhancement needs deeper neuroscience and better predictive models. The report attaches no dates to any of it.
The market has already moved. Consumer neurotechnology companies now outnumber medical ones, according to the Centre for Future Generations market atlas. Focus headbands are sold for office use. Organisations are trialling BCIs to monitor fatigue in aviation and mining.
Once a device becomes a condition of the job, the setting decides the ethics. Schools, workplaces and defence are named as the three environments marked by dependency or power imbalance. That yields a use-case test: which capability is affected, who benefits, who carries the risk, and whether refusal is genuinely possible.
Nine recommendations follow. Regulate by use case. Protect the right to refuse. Address equity early. Protect neural data. Promote reversibility. Require lifecycle accountability. Coordinate regulators nationally and internationally. Monitor technical and market signals. Consider social acceptability.
Pole two: the gap sits one layer below augmentation
Count what the report cannot show. No employer, school or military unit using BCIs for enhancement is named. No arrival year is given. No country-by-country rule comparison appears. Nine recommendations are addressed to a category that has not shipped.
The harms it fears are already deployed. Fatigue monitoring in aviation and mining is a live use, with an employer on one side of the neural data and a worker on the other. Coercion does not wait for superhuman capability. It arrives with a productivity dashboard.
That reading does not weaken the report. It relocates it. The binding constraints sit at the monitoring layer: who owns the signal, how long a device stays in a body, and who supports a user after the vendor exits. Those are procurement questions for this budget cycle.
Law is moving faster than the hardware. From 1 January 2027, California bars employers from using AI-enabled workplace tools to collect workers' neural data, with carve-outs for safety and some federally regulated defence and aerospace work. A UN Secretary-General report dated 27 August 2026 (A/81/351) recommended that states prohibit compulsory or coercive neurotechnology in the administration of justice. Both instruments regulate use. Neither needs an augmented human to exist.
The pattern is familiar. As we wrote in November, a defence neurotech startup flew three drones on EEG signals, then published no accuracy or latency data behind the demo. Capability claims arrive before the test file does.
Coercion, neural data and lifecycle accountability fall outside every current regime.
Defence buyers get a use-case test before the first augmentation procurement, not after.
What the 55 pages actually establish
The durable contribution is a continuum: treatment, restoration, assistance, optimisation, augmentation. One headset can sit at different points on that line depending on purpose, user and context. Device-class regulation keeps missing because the same object changes category as it moves.
| Setting | Rule already in force | What still falls through |
|---|---|---|
| Clinical | ✔ Medical device review, including transcranial stimulation with no therapeutic claim | ✗ Augmentation claims packaged as wellness |
| Workplace | ◐ California employer neural-data ban from 1 Jan 2027 | ✗ Optional-in-name-only productivity tools; cross-border arbitrage |
| Defence | ◐ Oversight and standards bodies named as coordination partners | ✗ No augmentation-specific standard; accountability after service |
Governing Augmentative Brain-Computer Interfaces, RAND Europe and Centre for Future Generations, October 2026
The first row carries more weight than the rest. The EU now classifies all transcranial brain stimulation devices as medical devices, requiring full Notified Body review even when no therapeutic claim is made. That closes one door on wellness hardware dressed as therapy. It leaves the same hardware dressed as productivity untouched.
Three issues are flagged as fitting no current framework: coercion in competitive settings, neural data governance, and accountability across a device's full life. Medical device rules handle safety. Data protection rules handle consent. Neither handles a worker who is told the headband is optional and knows the promotion is not.
Equity gets the same treatment. Anti-discrimination law does not yet draw a line between augmented and non-augmented people in a shared setting. The report asks for that line early, before access differences harden into a two-tier workforce.
Three recommendations that survive the framing
Three of the nine do not depend on augmentation arriving at all, which is why they carry the most weight for anyone buying or deploying a device.
The right to refuse is the first. In a hierarchy, consent is a schedule negotiation. California's new workplace rule already shows the shape of the compromise: a general ban on employer collection of neural data, then carve-outs for safety and for some federally regulated defence and aerospace work. Each carve-out is a place where refusal becomes a review board's problem.
Reversibility is the second. The report asks for substantially higher requirements on uses that are irreversible or hard to reverse, and adds a line that matters for procurement: reversibility alone is not grounds for approval. A removable sensor can still be a coercive one.
Lifecycle accountability is the third, and the most concrete. Long-term monitoring. Software updates. Device removal. Post-trial access. Cybersecurity. Support for users if the company leaves the market. Every item on that list is a line in a contract, and none of them appears in a medical device review that only asks whether a product is safe and does what it claims.
That is the report's quiet argument. Medical device regulation optimises a benefit-risk ratio. It does not price the worker who cannot say no, the operator whose implant outlives the vendor, or the school that adopts a headband because a neighbouring district did.
The paragraph with teeth for defence
The defence passage is short and concrete. Brain-machine interfaces could let personnel control multiple assets at once, improve learning and cognitive performance. Oversight bodies and standards organisations are listed among the regulators who must coordinate. No augmentation-specific standard is identified.
In defence, brain-machine interfaces could enable personnel to control multiple assets simultaneously, improve learning and cognitive performance.— Governing Augmentative Brain-Computer Interfaces, RAND Europe and Centre for Future Generations, October 2026
RAND's own 2020 assessment of military BCI is blunter. It named the trust deficit as the hard problem and called the technology a capability-vulnerability paradox: every gain in human-machine teaming opens an attack surface. Six years later the governance recommendation is still coordination.
Procurement teams can act before the standard exists. Ask at the trial stage who can refuse. Ask where neural data lives after the contract ends. Ask what happens to an operator when an implant cannot be removed and the vendor cannot be found. Lifecycle accountability is a contract clause, not a principle.
Decoders that work across users without retraining.
Implants that last longer inside the body.
Brain sensors built into earbuds, glasses and headsets.
Integration with operating systems and AI assistants, plus investment by large technology firms.
Trials in workplaces, schools and defence.
Where the two poles meet
Both readings survive contact with the evidence. The rules are early, and the gap they should close is already open. That combination is unusual, and it is the useful part.
The report says plainly that governance problems depend on how and where a device is used, not on whether the use counts as augmentation. Take that sentence seriously and the policy question changes. It stops being a debate about superhuman capability and becomes a procurement question about monitoring, consent and exit.
That question can be answered this year. The augmentation the 55 pages were written for is still a forecast.