The politics of the genome has changed. For years, arguments about genetic data were framed around a familiar medical scene: a patient in a clinic, a test ordered for a defined purpose, and a consent form meant to govern a bounded use. That picture is now inadequate. By 2026, human DNA is collected not only in hospitals and research programmes, but through consumer kits, fertility services, ancestry databases, pharmaceutical partnerships, forensic searches and large-scale biobanks. At the same time, machine-learning systems are finding patterns in genomic data that were not visible when many of those samples were first gathered. The result is a shift from episodic testing to continuous reuse. Genetic information is becoming long-lived digital infrastructure.
This matters because a genome is unlike most other forms of personal data. It is intimate, persistent and revealing, but also relational. It contains information about family members who never consented, about future disease risks that may never become illness, and about traits whose significance changes with scientific progress. A leaked bank card can be replaced. A compromised genome cannot. Nor is the problem confined to privacy in the narrow sense. In the agentic age, software systems can classify, predict, recommend and transact on the basis of biological signals at scale. Genetic rights therefore sit at the intersection of civil liberties, health governance, market structure and the political economy of AI.
Why genetic rights are becoming a first-order issue
Until recently, the genome sat at the edge of public policy: important, sensitive, but specialised. That is no longer tenable. AI systems trained on large biological datasets can support drug discovery, identify pathogenic variants, estimate disease risk and infer attributes from fragmented information. Even where genetics is only one feature among many, it can become a powerful anchor for identity resolution and predictive profiling. That makes questions of ownership and control more urgent than the older debate about whether individuals technically “own” their tissue sample.
The more practical question is this: who may decide how biological information is collected, linked, analysed, sold, licensed, retained and repurposed over time. Those decisions increasingly shape access to insurance, employment protections, reproductive choices, medical treatment and scientific progress. They also shape who captures economic value. If data is the feedstock of model development, then genomic datasets, and the rights attached to them, become strategic assets.
A genome is not a password that can be changed after a breach.
The old consent model is under strain
Modern bioethics rightly placed informed consent at the centre of genomic governance. Yet consent was designed for settings where the purpose of collection could be specified with some clarity. Broad consent in research, dynamic consent in digital studies and layered notices in consumer services all attempt to cope with uncertainty. But AI enlarges that uncertainty dramatically. Models can be trained years after data collection. New inferences can be drawn from old sequences. Linkage with other datasets can alter the sensitivity of information without any change to the underlying DNA.
This does not mean consent is obsolete. It means consent alone is insufficient. A person may agree to one use of a saliva sample and still have no meaningful understanding of downstream model training, cross-border data sharing, or probabilistic inferences generated later. Where power is asymmetric and technical consequences are opaque, formal permission can become a weak substitute for real control.
From one-time permission to ongoing governance
A more durable approach treats genetic rights as a matter of continuing governance rather than one-off authorisation. That implies constraints on secondary use, obligations of stewardship, robust audit trails and genuine withdrawal mechanisms where feasible. It also suggests that some uses should be prohibited even if users nominally agree to them. Democratic societies already take this view in other domains, recognising that contract cannot cure every imbalance.
Consumer genomics blurred the line between medicine and commerce
Direct-to-consumer genetic testing altered public expectations. It normalised the idea that DNA could be posted to a private company in return for ancestry insights, wellness claims or relative matching. This widened access and curiosity, but it also moved genomics into a commercial environment where terms of service, mergers, insolvency and advertising practices matter as much as medical confidentiality.
A genome is not a password that can be changed after a breach.
The central ambiguity has been persistent: are customers buying a test, joining a research endeavour, contributing to a data asset, or all three at once. In many cases, the answer has been some mixture. There is nothing inherently improper about commercial genomics, and some consumer databases have clearly aided family reunion, rare disease investigation and scientific work. But the governance challenge is obvious. If a firm changes strategy, is acquired, restructures or collapses, what happens to the biological information already collected and linked to family networks?
Such questions are no longer hypothetical. They expose a neglected point. Genetic rights must account not only for consent and privacy, but for corporate control, fiduciary duties and continuity of stewardship. Biological data can outlive the business model under which it was first surrendered.
Biobanks are public goods with private temptations
Large biobanks have become essential scientific infrastructure. Public and quasi-public programmes have shown the value of assembling genomic, phenotypic and longitudinal health data under governed access regimes. They can accelerate research, improve population health knowledge and support more representative science than fragmented small studies. In principle, a well-run biobank embodies social solidarity: individuals contribute data under rules meant to advance shared knowledge.
Yet success creates pressure. Valuable repositories attract commercial partnerships, geopolitical interest and demands for broader access. The line between public benefit and private extraction can become blurred, especially when downstream models or therapeutics generate outsized returns for a narrow set of actors. The legitimacy of biobanks depends less on rhetoric about innovation than on clear governance: transparent access criteria, participant representation, meaningful accountability and limits on uses that undermine trust.
There is also a representational question. Many genomic databases still over-represent certain populations. AI trained on skewed genetic data may perform unevenly across ancestry groups, exacerbating existing disparities in diagnosis and treatment. Here the language of inclusion must be handled carefully. More diverse data can improve scientific validity, but communities that have historically been exploited in research have reason to ask not only whether they are included, but on what terms and with what share of governance.
Genetic discrimination remains both narrower and broader than many assume
Public discussion of genetic discrimination often focuses on insurers or employers refusing opportunities on the basis of inherited risk. This remains a serious concern, and legal protections such as the United States Genetic Information Nondiscrimination Act address parts of it, though not comprehensively. Coverage differs across jurisdictions, and important gaps remain, particularly outside health insurance and employment, or where predictive analytics use proxies rather than explicit genetic fields.
But discrimination in the agentic age is broader than classical exclusion. It can include differential pricing, automated triage, risk scoring, reproductive pressure, targeted marketing and subtle forms of sorting that are difficult to detect. A system need not display a raw genotype to generate discriminatory outcomes. If AI models infer susceptibility, ancestry or familial traits from linked data, the formal absence of a genetic variable offers little reassurance.
Genetic data is never purely individual; it is shared, inferential and generational.
This is why governance must focus on effects as well as inputs. Anti-discrimination law built for overt classifications may struggle when statistical inference and behavioural proxies do the work indirectly. Genetic exceptionalism, meanwhile, has limits. DNA is not the only sensitive data. Yet neither should policy collapse it into generic personal information. Its permanence, familial nature and inferential richness justify special safeguards.
AI changes the stakes by turning data into reusable inference engines
The most consequential development is not simply the accumulation of genomic data, but the conversion of that data into models. A model trained on large-scale genetic and health datasets can encode relationships that persist even after access to the raw data is restricted. This complicates long-standing assumptions about anonymisation, deletion and withdrawal. If a participant leaves a study, can the effect of their data on a model be meaningfully removed? Sometimes perhaps, often not straightforwardly.
Genetic data is never purely individual; it is shared, inferential and generational.
The issue is not merely technical. It alters the location of power. Once value resides in a trained model, those who control model weights, evaluation protocols and deployment channels may command more leverage than those who hold original samples. The decisive struggle is shifting from access to repositories toward control over downstream inference systems. That includes models used in clinics, pharmaceutical research, reproductive services and consumer risk scoring.
Inference is the new frontier of ownership
Legal systems are generally better at governing collected data than generated inferences. Yet in practice, the inference may be more consequential than the sequence. A raw genome says little to most people; an AI-generated estimate of disease risk, likely drug response or familial relationship can shape real-world decisions. If firms or institutions claim broad rights over derived insights while individuals retain only nominal rights over source data, ownership becomes largely symbolic.
A serious framework for genetic rights must therefore address derived data, learned representations and model outputs. Otherwise, the protections of the sample stage will be bypassed at the inference stage.
Anonymisation is weaker than it looks
Genomic data has long challenged conventional anonymisation. A sufficiently rich sequence is inherently identifying in context, especially when linked with genealogical, demographic or health records. Research over many years has shown that de-identification is not a permanent state but a moving target, vulnerable to linkage attacks as more datasets become available. AI does not create this problem, but it sharpens it by improving pattern-matching and correlation at scale.
For policymakers, the lesson is sobriety. Promises that data is “anonymous” often overstate what can realistically be assured. Better practice is to discuss risk reduction, access control and governance rather than perfect anonymity. This is not pedantry. People may consent differently if they understand that biological re-identification is difficult to rule out, especially over long time horizons.
Property language is appealing but incomplete
Calls for individuals to own their genetic data are politically attractive. Ownership implies control, compensation and recourse. It speaks the language of sovereignty in a market society. There is merit in this instinct, particularly where people are asked to contribute valuable data while others capture most of the gains. Still, property metaphors can mislead.
A genome is not wholly alienable in the way a commodity is. It implicates relatives. It may bear on public health. Its meaning evolves with science. Purely proprietary regimes can also entrench inequality, favouring those best positioned to negotiate or monetise their data while doing little to restrain concentrated intermediaries. In some settings, a rights-and-duties model may be superior to a simple ownership model: individuals should have strong powers of access, portability, objection and redress, while stewards bear obligations of care, purpose limitation and accountable use.
Collective interests cannot be ignored
Because genetic information is shared across families and populations, governance cannot be reduced to atomised choice. Indigenous data governance debates have been especially instructive here, stressing that communities may hold legitimate interests in how biological data is collected and used. That does not negate individual autonomy; it places it within a wider social frame.
The market for genomic data needs rules before it deepens further
The decisive struggle is shifting from access to samples toward control over downstream models and inferences.
Data markets already exist in varied forms, from research access agreements to licensing arrangements and platform-mediated partnerships. Some exchange is valuable and necessary. Scientific progress depends on data sharing. Drug development would be slower without lawful, governed access to large datasets. The problem is not exchange as such, but asymmetry and opacity.
Many contributors do not know who ultimately uses their data, for what combination of research and commercial aims, or how long rights persist. Nor is it clear that one-off consent captures fair participation in value creation when AI systems can keep extracting utility from archived data. Policymakers should resist both naïve data mercantilism and simplistic anti-market reflexes. The task is to build conditions under which exchange is legible, constrained and socially legitimate.
That includes clear separation between clinical necessity and optional commercial use; transparent registries of approved access; independent oversight; and scrutiny of whether benefits flow back into public health, participants or neglected populations. In a sector built on trust, opacity is not merely an ethical weakness. It is a strategic liability.
What a durable settlement might look like
No single legal instrument will settle genetic rights globally. Jurisdictions differ, and health systems differ with them. But a coherent direction is visible. First, genetic data should be treated as highly sensitive by default, with strong restrictions on secondary use and international transfer unless equivalent safeguards apply. Second, rights must extend beyond collection to include derived inferences and model-mediated decisions where these materially affect people.
Third, institutions holding genomic data should carry heightened stewardship duties. In practice that means minimisation, security, governance transparency, documented access decisions and mechanisms for participant challenge. Fourth, anti-discrimination frameworks must be updated for inferential systems, not just explicit genetic fields. Fifth, public-interest research should be protected, but not with a blank cheque. Trustworthy science requires legitimacy as well as access.
Technical measures matter too. Privacy-preserving computation, secure research environments, provenance tracking and auditable model development can reduce risk, though none is a panacea. Their value lies in shrinking the gap between formal promises and actual practice.
Sovereignty in biology will not mean isolation
There is a temptation, faced with these risks, to conclude that the safest genome is the one never shared. For some individuals that will be a rational choice. Yet total enclosure has costs. Genomic research has contributed to diagnosis, rare disease discovery and the development of more precise therapies. The goal should not be autarky in biological information. It should be governed participation on fair terms.
That is the broader lesson for the agentic age. Sovereignty is not the same as withdrawal. It means retaining meaningful power over how one is represented, analysed and acted upon by systems that can outlast the original transaction. In genetics, that principle is especially urgent because biological information is durable, networked and generational. It binds the intimate to the infrastructural.
The foundational question for this category, then, is not simply whether DNA should be private. It is how democratic societies should allocate authority over the most personal dataset humans possess once it becomes fuel for autonomous inference. If governance remains stuck at the point of collection, while value and power move downstream into models, contracts and data markets, individuals will formally consent to a system they do not in substance control. Genetic rights worthy of the name must govern the full lifecycle of biological information, from sample to sequence to model to decision.
The decisive struggle is shifting from access to samples toward control over downstream models and inferences.
That struggle will shape far more than privacy law. It will influence the legitimacy of precision medicine, the architecture of health-data markets, the boundaries of insurance and employment fairness, and the extent to which ordinary people retain sovereignty over their biological identity. In the coming years, the genome will be contested not only in laboratories and legislatures, but in cloud infrastructure, procurement frameworks and machine-learning pipelines. That is why genetic rights and ownership now sit at the centre of technological self-government.
