The European Group on Ethics in Science and New Technologies wants the EU to regulate the emerging combination of neuroscience and artificial intelligence as an infrastructure problem, arguing that rules focused on individual devices or isolated datasets will struggle once neurological information begins moving through reusable models and larger technical systems.

Its statement, Governing Neuro-AI: Towards an Infrastructure Approach, calls for stronger protection of neurodata and information inferred from it, safeguards around Brain Foundation Models, controls where neurological information influences consequential decisions, and greater European capacity to govern the systems through which those technologies are developed and deployed.

The group also wants a targeted fitness check of the existing EU regulatory framework rather than assuming that data protection, medical-device law, research governance, and AI regulation already fit together cleanly. Neuro-AI sits across those regimes because the same dataset can move from research into model development and then towards applications whose commercial or public-service use was not envisaged when the information was first collected.

That shift changes the regulatory problem. A rule governing whether one organisation can collect one piece of sensitive information is only part of the picture once large datasets can be combined, reused, and transformed into models capable of generating new inferences about people.

Neurodata moves beyond the device

Neurotechnology covers systems that measure or interact with activity associated with the nervous system, while advances in AI make it possible to analyse patterns across increasingly large and complicated datasets. When the two converge, the output can extend beyond the raw signals recorded by a device because a model may infer characteristics, conditions, or behavioural patterns that were not directly measured.

That distinction is commercially significant because modern AI development rewards reuse. A dataset gathered for healthcare or research can acquire additional value if it later contributes to model development, benchmarking, or other services, but the people represented in the data may have limited visibility into those later uses and little practical ability to contest what a model infers.

The EGE’s focus on Brain Foundation Models pushes the issue further upstream. Foundation models derive much of their utility from broad training data and the ability to support several downstream tasks, so neurological versions could place considerable technical and commercial power in the hands of organisations able to assemble sufficient data, compute, scientific expertise, and distribution.

Healthcare provides the most obvious route into deployment, although the implications are not confined to clinical settings. Employers, insurers, researchers, public bodies, and technology suppliers could all encounter Neuro-AI through different applications, while information generated in one environment may eventually influence decisions in another if governance fails to follow the data and model.

Sovereignty enters a more sensitive layer

The EGE’s infrastructure framing also places Neuro-AI inside Europe’s wider debate over control of strategic technology. Cloud concentration, semiconductor capacity, AI models, and data access have all become industrial-policy concerns because dependency at one technical layer can limit choices throughout the rest of the stack.

Neurological information makes that dependency more sensitive. A conventional business dataset can reveal transactions, preferences, or performance, whereas neurodata and derived inferences may expose characteristics that individuals cannot easily observe or verify themselves, deepening the asymmetry between the organisation operating the system and the person represented within it.

If valuable neurological datasets and models become concentrated in infrastructure controlled outside Europe, familiar sovereignty questions would reappear around research access, procurement, commercial dependency, and where sensitive information is processed. The difference is that the underlying data may reveal a more intimate layer of information than most enterprise or consumer datasets currently under discussion.

The group’s response is not to demand a single new Neuro-AI law immediately, but to examine whether existing rules leave important gaps once technologies are treated as connected infrastructure. That approach gives policymakers room to test how the AI Act, GDPR, medical-device rules, research frameworks, and other legislation interact before deciding whether additional measures are required.

Regulation arrives before market lock-in

The EGE is an advisory body rather than an enforcement authority, so its statement creates no direct obligations for companies or public bodies. Its recommendations can nevertheless shape Commission thinking, particularly because the group is asking regulators to address technical architecture while much of the field is still research-led rather than waiting for a mature market to form around a handful of platforms.

That timing gives Europe an opportunity that has often been missing in digital regulation. Rules governing social platforms, online advertising, and app stores largely arrived after powerful commercial structures were already established, leaving regulators to intervene in markets where business models, customer behaviour, and technical standards had become difficult to unwind.

Neuro-AI is earlier in its development, although that does not make regulation straightforward. Policymakers still have to avoid constraining legitimate medical research or useful assistive technologies while creating boundaries strong enough to prevent neurological information becoming another broadly reusable dataset whose downstream applications are difficult to control.

The EGE’s infrastructure approach consequently shifts the debate away from speculative images of mind-reading devices and towards the less theatrical systems that determine where data sits, how models are trained, who can reuse them, and which organisations control the technical capacity. Those choices are likely to shape Neuro-AI long before the technology becomes routine enough to attract mass-market attention.