Ethics in Cognitive Enhancement
Linas JuozėnasShare
Ethics in cognitive enhancement
Expand what people can do. Protect their freedom to decide.
The wish to learn, remember, communicate and think more clearly is deeply understandable. Better tools may open opportunities for study, creative work, independence and participation. Whether an intervention deserves a place in someone's life depends on more than the promise of improved performance.
What benefit has actually been demonstrated? Does the person want it? Can they decline without losing opportunities? Who controls the information collected, and who remains responsible if something goes wrong?
These questions make ethics part of good design and good decision-making. This guide explores them through evidence, practical examples and the European and international frameworks that help protect people as technology develops.
Ask what they want to make possible, and whether the proposed tool serves that aim.
A demonstration, a product claim and a reliable everyday benefit are different things.
Users should have choices, while developers and institutions remain accountable for their decisions.
What do we mean by cognitive enhancement?
Cognitive enhancement broadly concerns attempts to improve capacities such as learning, attention, memory or reasoning. Discussions often focus on uses beyond treating an illness, but the boundary between treatment, assistance, restoration and enhancement is not always clear.
A communication aid can help someone express abilities they already possess. A learning tool can make information easier to use. A proposed biological intervention may aim to change a cognitive process itself. Describing the intended purpose is more useful than assuming these are equivalent.
Familiar aims, different responsibilities
Books, teaching and external memory aids already extend what people can accomplish. New approaches can introduce additional questions about bodily intervention, sensitive information, dependence on a supplier or effects that are difficult to undo.
| Approach | What is being proposed? | A central ethical question |
|---|---|---|
| Learning and digital support | Practice, feedback, reminders, accessible information or assistance with a task. | Does it build useful capability, and can the user evaluate its output? |
| Biological interventions | A substance or procedure intended to change a bodily process relevant to cognition. | Is there convincing evidence for this purpose and this person, with acceptable burdens? |
| Neurotechnology | Recording or influencing nervous-system activity, including brain stimulation and brain–computer interfaces.[1] | Who controls the intervention, its settings and the information it produces? |
| Genetic proposals | Changing DNA, potentially with consequences extending beyond the individual. | What can realistically be predicted, and who will bear the consequences? |
These categories describe approaches, not endorsements or a ranking of effectiveness. The rest of the guide asks how to assess a specific proposal in its actual setting.
What has improved, for whom, and for how long?
“Improves cognition” can conceal several different claims. Becoming faster at a practised game, remembering a particular word list and making better decisions in everyday life are different outcomes. Evidence for one should stay attached to that outcome.
A 2016 review of brain-training claims found strong support for improvement on practised tasks, with much weaker support for broader transfer. The journal's publisher describes problems including inadequate comparisons and claims extended beyond the populations studied.[2] This historical example illustrates how to examine a claim; it is not a verdict on every present or future product.
Ask for the comparison that matters
Would the result also occur with ordinary practice, attention from an instructor or a simpler tool? Did the study compare similar experiences? Were the outcomes chosen before the results were known? Were disappointing findings reported?
A useful account names the participants, comparison, duration, measured benefit and uncertainties. It also explains whether people continued to benefit outside the research setting.
Keep a demonstration in proportion
A successful laboratory demonstration can justify further research. It does not automatically establish a dependable consumer service, a lasting increase in general intelligence or a suitable intervention for children.
Honest uncertainty can help a person decide whether the opportunity interests them. Inflated claims make that decision harder, even when the underlying research is valuable.
Autonomy, benefit, care and fairness work together
UNESCO's Universal Declaration on Bioethics and Human Rights connects respect for dignity and autonomy with benefit, harm reduction, justice and protection against discrimination.[3] These principles can guide questions about a proposal without producing the same answer in every case.
Autonomy
Does the person understand the choice and have meaningful room to make it? Their aims and preferences should shape what happens.
Benefit
What improvement would matter in their life? A higher score has value only in relation to a worthwhile purpose.
Avoiding harm
What burdens could follow, and how can they be reduced? Include health, privacy, money and the ability to change course.
Justice
Who can access the benefit, who carries the risks, and who might be excluded by a new expectation?
Autonomy includes the choice to pursue improvement
A person may reasonably value an opportunity that others would decline. Respect involves listening to that aim, giving accurate information and making responsible pathways available where possible.
Equally, choosing a different life should remain possible. A student may value learning without constant measurement; a worker may prefer a role that does not demand bodily intervention. Ambition takes many forms.
When values conflict, explain the conflict. For example, a less invasive option may offer a smaller benefit but fit someone's priorities better. Another person may make a different choice within the same lawful, ethically acceptable options.
Informed consent is a process of understanding and choice
A signature records a decision; it does not demonstrate understanding by itself. The Oviedo Convention's general rule for health interventions requires free, informed consent after appropriate information about the intervention and its consequences and risks. It also provides for withdrawal.[4]
Explain the actual offer
Someone considering a proposal should be able to understand its purpose, what they would experience, the evidence for benefit, foreseeable burdens, important unknowns, available alternatives and the practical consequences of stopping.
Information should be usable: plain language, suitable communication support, time for questions and an opportunity to discuss the decision privately. Asking someone to explain the offer in their own words can reveal where the explanation needs improvement.
Distinguish the decisions being made
Agreeing to use a device, joining research, sharing identifiable data and permitting promotional use of a photograph are distinct decisions. Present them clearly enough that a person can understand which choices are necessary for the service and which are optional.
Return to the conversation when something important changes
New safety information, a materially changed purpose or a substantial change in what the system does can call for another discussion and, where required, renewed consent. Do not bury a meaningful change inside routine paperwork.
Withdrawing permission cannot always reverse a completed procedure or remove every lawful research record. Explain those limits beforehand. An honest description of stopping is more useful than an assurance that everything is reversible.
A choice can be pressured even when it is called voluntary
Consider what happens after someone says no. Will their grades, pay, opportunities or relationships suffer? Does a supervisor receive the list of people who declined? Can a person ask questions without being labelled uncooperative?
In data protection, the European Data Protection Board warns that dependence in employment generally makes consent an unsuitable basis for workplace processing. Exceptional situations require genuinely free choice without adverse consequences.[5] The broader ethical lesson is to examine the conditions around an agreement.
Education
A learning opportunity should offer accessible participation and clear assessment rules. Students need room to raise concerns without risking a teacher's goodwill or their place in a programme.
Employment
Ask whether the proposal serves workers' interests as well as productivity. Workload, staffing and breaks deserve attention before a bodily intervention becomes an expected solution.
Hierarchical professions
Where authority is strong, independent advice and review become especially important. Research participation, operational duties and clinical care should be distinguished clearly.
Build a credible alternative
An institution can offer an equivalent route, keep participation decisions away from performance assessment, and provide a confidential channel for concerns. The arrangement must work in practice; a promise of voluntariness means little if declining quietly closes opportunities.
Payment is not automatically coercive. Reimbursing time, travel or inconvenience may make participation fairer. The question is whether the overall offer obscures risks or exploits a person's circumstances.
Children and people needing decision support deserve a voice
Difficulty communicating should prompt an effort to make the conversation accessible. It should not lead directly to an assumption that the person has nothing meaningful to say about their own life.
Where a minor cannot legally consent to a health intervention, the Oviedo Convention provides for appropriate authorisation and increasing weight for the child's views with age and maturity. It also calls for involvement, as far as possible, of adults who cannot consent themselves.[4]
Ask whose goal is being pursued
A child's wish to communicate more comfortably may differ from an adult's wish for higher marks or quieter behaviour. Make that difference visible. Consider how an intervention might affect the child's daily experience, relationships and future options.
When discussing an optional enhancement, asking what could be gained by waiting can be as useful as asking what could be gained now. The answer depends on the proposal and its evidence; waiting is not automatically the right choice for needed care.
Avoid both exclusion and unnecessary exposure
Protecting people should not mean excluding them from every opportunity to benefit from research. Inclusion needs a defensible purpose, suitable safeguards and attention to the person's preferences. A diagnosis or disability does not tell us whether an individual wants a particular technology.
For families and professionals, a useful question is: “How can this decision preserve the person's interests and participation, including the person they are becoming?”
A worthwhile benefit must be considered alongside its burdens
Risk is specific to an intervention, its use and the person affected. A reminder app, externally applied stimulation and an implanted system require very different assessments. A reassuring description such as “personalised” or “non-invasive” cannot substitute for evidence.
Look beyond the headline outcome
Ask about discomfort, unwanted effects, follow-up, maintenance and what support would be available if the intervention no longer suited the person. Where effects are uncertain or difficult to reverse, that uncertainty belongs near the centre of the decision.
Also include time, travel, training, subscriptions, help from family and the effort of using the system. A technically successful intervention may still demand more than someone wants to take on.
Compare with the available alternatives
Could a simpler tool, an accessible workplace, different teaching or appropriate care meet the same need? Sometimes the new intervention offers something distinctive. Sometimes changing the environment answers the problem more directly.
This comparison should consider the burden of doing nothing as well. For someone facing a substantial limitation, preserving the status quo can carry a real cost.
Use a proportionate approach
Greater potential benefit can justify investigating a more demanding intervention under suitable conditions. Strong enthusiasm cannot make poor evidence stronger, and a person's agreement does not remove the responsibilities of those offering it.
The OECD's responsible-innovation framework connects safety assessment with inclusion, oversight and protection of brain data.[6] Considering these together helps prevent a narrow performance measure from dominating the whole decision.
Mental privacy concerns both data and the claims made from it
Signals recorded from the nervous system can be sensitive. So can conclusions drawn from them. An inaccurate label about someone's attention or emotional state can still cause harm if another person treats it as authoritative.
Keep technical capability in perspective
In a 2023 study, researchers reconstructed aspects of language meaning from functional MRI recordings. Their successful decoding required participant cooperation during both training and use.[7] This was a specific research system, not evidence of unrestricted access to anyone's private thoughts.
It is possible to take privacy seriously while describing capabilities accurately. A prediction should be presented with its limitations, especially when it could affect someone's opportunities or reputation.
Understand what information leaves the device
Ask what is collected, what is inferred, who receives it, why it is needed and how long it is kept. Distinguish information needed to provide the service from optional research or commercial uses. Consider whether processing could happen with less identifiable information.
Under the GDPR, personal data need a lawful basis and protections including purpose limitation, data minimisation and appropriate security. Health and genetic data, and biometrics used for unique identification, fall within Article 9's specially protected categories. The exact treatment of a neural dataset depends on its content and use.[8]
Consent alone does not excuse unnecessary or unfair processing, and it is one of several possible legal bases.[5] Organisations must explain their basis and fulfil the associated duties.
Identity and agency are practical concerns, not just philosophical puzzles
A person may experience assistance as a welcome expression of who they are. Another may feel uncomfortable with a tool that changes their routines or the way they communicate. Those experiences should be heard without imposing one account of an “authentic” self.
Ask who remains in control
For a system that suggests words or actions, can the person inspect, reject or correct the output? For a system that adjusts itself, are meaningful changes explained and appropriate controls available? What happens when the person and the system disagree?
These questions become particularly important when output is attributed to the user. A predicted sentence should not automatically be treated as an expression of someone's beliefs or intentions.
Allow differences in valued ways of thinking
People may prioritise creativity, careful reflection, emotional richness, communication or speed differently. An institution's preferred performance measure should not become the only acceptable model of a mind.
Respect for cognitive diversity is compatible with helping a person change something they find difficult. The person's own aims matter. Offering assistance and accepting a refusal can both express respect.
Discuss authenticity without policing it
One ethical concern is that enhancement could distance achievement from effort or valued experience. A different view is that tools have always helped people express their abilities. Neither argument settles every case. Ask what the activity is for, what others reasonably expect, and whether the person endorses the change.
Fairness includes access, accommodation and freedom to decline
A tool could reduce a barrier for one person while becoming an expensive expectation for everyone else. Ethical assessment should examine both possibilities rather than assuming innovation must either widen or close inequality.
Count the full cost of access
Price is only part of the picture. Consider language, accessible instructions, transport, reliable connectivity, repairs and continuing support. A service that exists only near a specialist centre may be difficult to use from a small Lithuanian town, even when the initial device is funded.
Designers can consider remote support where appropriate, local partnerships, usable documentation and funding that covers maintenance. These choices can be discussed early, while the service is still taking shape.
Match fairness to the activity
In an examination, fairness depends partly on what the assessment is intended to measure and which forms of assistance are allowed. In everyday work, a helpful tool may expand participation without undermining the purpose of the task.
Accommodation for disability should not be casually equated with cheating. Rules need a clear purpose and a way to consider different needs.
Watch for a rising expectation
If an optional enhancement becomes necessary to remain employable or competitive, access alone will not resolve the pressure. Institutions may need to reconsider performance expectations and protect meaningful alternatives.
Public investment also involves choices. A proposal for expensive enhancement should be considered alongside education, accessible environments and needed care, with reasons that affected communities can examine.
Genetic enhancement raises questions that extend beyond one decision
Intelligence is a complex trait shaped by genetic and environmental influences. MedlinePlus Genetics describes contributions from many genes alongside learning resources, health and other developmental conditions.[9] An association with a measured trait does not by itself identify an edit that would safely improve a person's future abilities.
Distinguish changes to an individual from heritable changes
Somatic interventions concern cells of the person treated. Heritable changes can also affect descendants. WHO's 2021 genome-editing work addresses these differences through recommendations on oversight, registries, international cooperation and public participation.[10]
Parents make decisions, but not without limits
Children cannot consent to many decisions that shape their early lives. That fact alone does not settle which decisions are acceptable. Relevant questions include likely benefit, uncertainty, alternatives, reversibility and how much the decision constrains the child's future.
When possible consequences extend across generations, individual parental agreement cannot answer every public question. There are interests beyond the immediate family and uncertainties no consent form can remove.
Keep law attached to its jurisdiction
For states bound by the Oviedo Convention, Article 13 limits genome-modifying interventions to preventive, diagnostic or therapeutic purposes and excludes an aim of changing descendants' genomes.[11] This is a specific treaty provision, not a description of one uniform worldwide law.
Ethical foresight can explore future possibilities without presenting speculative intelligence enhancement as an established clinical service.
European law and international guidance do different jobs
There is no single rule covering everything described as cognitive enhancement. The relevant framework depends on the product, purpose, setting, data and jurisdiction. Lithuania ratified the Oviedo Convention in 2002; it entered into force for Lithuania in 2003.[12]
The table identifies selected rules and their scope. It is an orientation to the issues, not a complete assessment of a particular product or service.
| Framework | Status and relevant focus | A limit to remember |
|---|---|---|
| Oviedo Convention | A binding treaty for its parties, addressing human rights in biomedicine, consent and genetic interventions.[4] | It is a Council of Europe treaty; participation is not identical to EU membership. |
| EU General Data Protection Regulation | Binding rules for personal-data processing within its scope, including additional conditions for special categories.[8] | A data-protection basis does not authorise a bodily intervention. |
| EU Medical Device Regulation | Governs medical devices and certain products without a medical purpose. Annex XVI includes specified brain-stimulation equipment.[13] | Scope, intended purpose and classification need assessment; “neurotechnology” is not one uniform device class. |
| EU Artificial Intelligence Act | Article 5(1)(f) prohibits specified AI use to infer emotions in workplaces and education, with an exception for medical or safety purposes.[14] | This is not a blanket ban on all physiological monitoring. The system and its intended use matter. |
| UNESCO's 2025 neurotechnology recommendation | International guidance on rights, autonomy, mental privacy and responsible development, adopted on 11 November 2025.[1] | A recommendation is not itself a directly enforceable worldwide product law. |
| OECD responsible-innovation recommendation | A framework for governments and innovators covering safety, inclusion, oversight, data and societal dialogue.[6] | Practical implementation and enforceable duties depend on the relevant institutions and laws. |
Standards can guide practice without being statutes
Research declarations, professional codes and international recommendations can influence ethics review, funding and institutional policy. Their authority and enforcement mechanisms differ from legislation.
Cooperation matters across borders
Research, services and data can cross jurisdictions. Responsible providers need to account for the relevant countries and arrangements. Offering a service elsewhere does not resolve questions about evidence, consent or responsibility for follow-up.
Legal provisions are checked here as of 5 September 2026. Later changes, national requirements and the details of a proposed use can change the assessment.
Good governance makes valuable progress easier to trust
A useful approach combines early learning with safeguards proportionate to the consequences. Low-burden design work can involve users early. A bodily intervention or consequential use of personal data needs the appropriate scientific, ethical and legal foundations before people are exposed to it.
Give affected people influence
Ask potential users what would count as benefit, what would make participation difficult and what forms of support they would need. Include people who might decline the technology; their concerns can reveal design problems that enthusiasm misses.
Make research accountable
The 2024 Declaration of Helsinki calls for independent ethics review, transparent research registration and reporting, protection of participants and arrangements for appropriate post-trial access. It also makes clear that participant consent does not transfer researchers' protective responsibilities to participants.[15]
Scientific freedom benefits from a process that produces credible evidence. Publishing negative results and explaining conflicts of interest help others decide what a finding actually supports.
Plan for ordinary failures and the end of support
Who handles repairs, software changes, discontinuation or an organisation closing? What would someone need to move to another service? For an implanted system, what clinical support and options would remain available?
These questions belong in planning, purchasing and research discussions. A long-term promise needs identifiable responsibilities and resources behind it.
Leave room to revise the decision
Set clear conditions for reviewing benefit, investigating problems and changing or ending a programme. Good governance can permit useful development while remaining responsive to what people experience.
Turn an attractive proposal into an answerable decision
The following scenarios are fictional illustrations. They show how a change in purpose, power or support can change the ethical assessment without requiring a different technology.
A study-support service
A college offers an optional programme and describes exactly which skills it practises. Students can use an equivalent route, and participation does not affect grades.
Questions remain about evidence, access and data. The credible alternative strengthens the basis for a voluntary choice.
A workplace headset
A manager offers a focus device, receives individual scores and begins using participation in promotion discussions.
The priority is to reconsider that arrangement: evaluate the claims, remove pressure, assess lawful data use and involve workers in the decision.
An experimental interface
A research team describes uncertain benefits and continuing technical support, but cannot explain what happens when funding ends.
The missing support plan is part of the offer itself. It needs a concrete answer before a person can weigh the commitment.
Five steps for a clearer conversation
-
Name the person's purpose.
Describe the practical capability or opportunity they want. Ask who proposed the goal and whether the person shares it.
-
Check the evidence and alternatives.
Identify what has been demonstrated, in which people and settings, and what remains uncertain. Compare with other ways to meet the need.
-
Examine the freedom to choose.
Consider power, payment, privacy and the consequences of saying no. Make room for accessible explanation and independent advice.
-
Account for the whole commitment.
Include data, cost, burdens, follow-up, control, changes and stopping. Identify who is responsible for each important part.
-
Record the reasons and review conditions.
A decision may be to proceed through an appropriate pathway, revise the proposal, seek more evidence or decline. State what would justify reconsidering it.
The aim is a decision that can be understood and questioned by the people affected. A careful process should make their options clearer.
Answers to recurring ethical questions
Is cognitive enhancement inherently unethical?
No single answer fits every intervention. Consider its purpose, evidence, burdens, voluntariness and effects on others. A tool can expand someone's opportunities while still requiring clear limits and responsibilities.
Does a signed consent form make an intervention ethical?
No. Understanding and freedom matter, and consent does not remove duties concerning evidence, safety, privacy or fair treatment. Some proposals remain unacceptable or unlawful even when a person agrees.
Can someone freely choose to take a risk?
In appropriate circumstances, an informed adult may accept risks others would decline. Respect for that choice belongs alongside safeguards, applicable law and the responsibilities of those offering the intervention.
Does “non-invasive” mean harmless?
No. It describes a feature of how an intervention is delivered. Its effects, suitability, evidence and other burdens still need assessment. Privacy and financial consequences also deserve attention.
Can brain technology read anyone's thoughts?
That sweeping claim is unsupported by the research example discussed here. The 2023 language-decoding study used a specific fMRI system and required cooperation.[7] Evaluate claims about another system on its own evidence.
Are all neural data automatically health data under the GDPR?
Classification depends on what the data reveal and how they are used. Personal neural data can raise Article 9 issues, but the label “neural” alone does not settle every legal question.[8]
Can an employer make monitoring acceptable by calling it optional?
The actual consequences of refusal matter. Employment power imbalances make reliance on consent problematic in many situations.[5] A label cannot replace a proper assessment of the arrangement and applicable law.
Is using assistance the same as cheating?
No. It depends on the activity's purpose and rules. A disability accommodation, an everyday work tool and unauthorised assistance in an examination need different assessments. Clear, fair rules help everyone understand expectations.
Does UNESCO's recommendation create a worldwide ban?
No. It provides international ethical guidance. Enforceable requirements come through the relevant legal frameworks and their implementation. The distinction between guidance and binding law is explained in section 12.
What should remain central as technologies change?
The person's own aims, honest evidence, meaningful choice and accountable support. New capabilities can be valuable while people retain the freedom to question them, help shape them or choose another path.
Sources, legal scope and the limits of examples
Sources checked: 5 September 2026. Scientific findings, ethical arguments and legal requirements are distinguished throughout. The original research example illustrates a particular capability under studied conditions. Ethical scenarios and decision questions are this guide's practical illustrations.
The legal discussion identifies selected provisions rather than every applicable rule. Treaty participation, product classification, national law and the specific use all matter. The AI Act and Medical Device Regulation were checked against the consolidated texts identified below; the underlying Official Journal acts remain the authoritative legal texts.
- UNESCO. Recommendation on the Ethics of Neurotechnology. Adopted 11 November 2025.International ethical recommendation addressing the technology's whole life cycle. A normative framework, not a directly enforceable worldwide product statute. Return to text ↑
- Association for Psychological Science. Brain-Training Claims Not Backed by Science, Report Shows. Publisher's account of the Simons and colleagues review, 3 October 2016.Reports the review Do ‘Brain-Training’ Programs Work? Used as a historical example of transfer and study-design issues, not a current review of every product. Return to text ↑
- UNESCO. Universal Declaration on Bioethics and Human Rights. Adopted 19 October 2005.Ethical framework concerning dignity, autonomy, benefit, harm, justice and discrimination. The article's examples apply these ideas; they are not quotations or legal determinations. Return to text ↑
- Council of Europe. Convention on Human Rights and Biomedicine (Oviedo Convention). Treaty ETS No. 164, 4 April 1997.Treaty text. Articles 5–6 address consent and participation; Article 13 concerns genome modification. Binding obligations apply to its parties within the treaty's scope. Return to text ↑
- European Data Protection Board. Guidelines 05/2020 on consent under Regulation 2016/679. Adopted 4 May 2020; version 1.1.Official GDPR interpretation, including employment power imbalances and genuine choice. Data-processing consent and consent to an intervention are distinct. Return to text ↑
- Organisation for Economic Co-operation and Development. Neurotechnology. Policy overview and responsible-innovation resources.Explains the OECD approach and links to its 2019 recommendation. Covers safety, rights, inclusion, data protection, public dialogue and international cooperation. Return to text ↑
- Tang J, LeBel A, Jain S and Huth AG. Semantic reconstruction of continuous language from non-invasive brain recordings. Nature Neuroscience, 2023; 26:858–866.Original experimental research using fMRI. Cooperation was required to train and apply this decoder. The finding does not establish unrestricted thought-reading capabilities. Return to text ↑
- European Parliament and Council of the European Union. General Data Protection Regulation — Regulation (EU) 2016/679. 27 April 2016; selected provisions checked.Articles 4–9 and related provisions inform the discussion of personal data, processing principles, lawful bases and special categories. Dataset-specific assessment remains necessary. Return to text ↑
- MedlinePlus Genetics, US National Library of Medicine. Is intelligence determined by genetics?. Updated 1 August 2023.Scientific public-information overview of genetic and environmental influences. Associations with a complex trait do not establish a safe enhancement intervention. Return to text ↑
- World Health Organization. WHO issues new recommendations on human genome editing for the advancement of public health. 12 July 2021.Official account of WHO's governance framework and recommendations. Distinguishes somatic, germline and heritable editing; international recommendations do not create uniform national law. Return to text ↑
- Council of Europe. Genome editing technologies: final conclusions of the re-examination of Article 13 of the Oviedo Convention. 10 October 2022.Official clarification of the permitted purposes of genome interventions and the prohibition on an aim of modifying descendants' genomes under this treaty. Return to text ↑
- United Nations Treaty Collection. Convention on Human Rights and Biomedicine: treaty registration and participation. Treaty participation record.Records Lithuanian ratification on 17 October 2002 and entry into force on 1 February 2003. Used to establish the treaty's relevance to Lithuania. Return to text ↑
- European Parliament and Council of the European Union. Medical Device Regulation — Regulation (EU) 2017/745. Consolidated text dated 19 July 2026.Selected scope and classification provisions and Annex XVI, point 6. Specific products need assessment; consolidated texts are documentation tools linked to the authoritative legal acts. Return to text ↑
- European Parliament and Council of the European Union. Artificial Intelligence Act — Regulation (EU) 2024/1689. Consolidated text dated 27 July 2026.Article 5(1)(f) addresses emotion inference in work and education, with medical or safety exceptions. This prohibition has applied since 2 February 2025. Return to text ↑
- World Medical Association. Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Participants. Revised 19 October 2024.Professional research-ethics declaration. Relevant provisions address independent review, consent, reporting, responsibility and post-trial arrangements; it is not itself a worldwide statute. Return to text ↑