Abbott and Google want your glucose data. But who benefits most?

You're probably familiar with buying a glucose sensor, or getting one on prescription, attaching it to your own body and generating thousands of readings day and night. Those readings tell you something about your health - your glucose levels - but they could also tell you much more. Combined with information from thousands of other people, they could become considerably more valuable. They can become the raw material used to develop artificial intelligence capable of interpreting human health.

And that is where a new partnership between Abbott and Google starts to raise significant questions about who benefits from our health data.

The two companies have just announced plans to bring data from Abbott's Lingo glucose biosensor into Google Health, where glucose trends can be combined with information about meals, exercise, sleep and other aspects of everyday life to provide more personalised health guidance. Google says full product details, integrations and availability will follow later this year. 

But Abbott and Google aren't stopping at personalised advice. Abbott has also announced what it describes as one of the largest real-world metabolic health studies to date, combining continuous glucose readings with wearable, laboratory and survey data. The research is intended to contribute to future AI-powered guidance and metabolic-health innovation.

Google is already moving further in this direction.

Just a day after Abbott and Google announced their partnership, Google unveiled a new Pixel Watch and Fitbit feature capable of tracking trends associated with insulin resistance without measuring glucose or taking a blood sample. The system uses AI to analyse physiological information collected passively by the wearable over several weeks and is designed to identify changes in metabolic health that might otherwise remain unnoticed.

The technology behind it draws on AI models trained using enormous quantities of information generated by wearable users. In other words, health data generated across huge populations is already contributing to AI capable of extracting entirely new health insights from signals those devices were collecting.

That makes the questions raised in this article considerably less theoretical.

That creates an extraordinary exchange. Consumers buy the sensors and wearables, their bodies generate the biological information and their everyday lives provide the behavioural context. The resulting research can then contribute to AI, algorithms and future commercial health products controlled by some of the world's largest medtech and technology companies.

The individual doesn't own the AI that results. If these systems become increasingly important in interpreting our health, the companies controlling them could eventually control something considerably more valuable than the sensor: access to the knowledge generated from our collective health data.

The concern is therefore no longer simply about privacy, or even who profits from your glucose readings. It is whether consumers are providing the raw material for a new generation of proprietary health intelligence that society could ultimately become dependent upon.

You buy the sensor. Your body supplies the data

The commercial model behind consumer health technology is worth considering because it turns some of the assumptions around traditional medical research on their head.

A medical-device company conducting conventional clinical research may need to recruit participants, supply the equipment, arrange testing and pay the considerable costs involved in collecting reliable health information.

Consumer wearables have already created much of that infrastructure. People buy glucose sensors, smartwatches and smart rings because they want information about their own health. They then wear those devices while eating, sleeping, exercising and working, producing an extraordinarily detailed record of everyday human biology.

For the individual, the attraction is personal insight. Across sufficiently large populations, however, those measurements become an enormous research resource. But this isn't a resource generally shared with independent scientists, clinicians, medical researchers or medical peers. Abbott and Google control their respective commercial data environments, including how access to the information is provided and how knowledge generated within those environments is developed.

That matters because publishing selected research findings or submitting a paper for peer review is very different from giving independent researchers access to the underlying data. Without equivalent access, scientists outside the companies cannot freely interrogate the same enormous dataset, ask different questions of it, challenge how relationships have been interpreted or attempt to reproduce findings using the original information.

Google's own Health API research policy illustrates the distinction. It requires qualifying health research to undergo ethical oversight and places restrictions on how participant information can be transferred and used. That does not turn the resulting commercial dataset into an open scientific resource.

The result is an unusual concentration of scientific power. Consumers generate the biological data, but access to the resulting research resource can remain controlled by the companies collecting it. Data capable of advancing wider understanding of metabolic health can therefore also become a proprietary advantage, feeding commercial AI and product development rather than becoming an equivalent research resource for the wider medical and scientific community.

The Abbott-Google study plans to combine continuous glucose data with wearable information, laboratory results and surveys. That potentially provides the companies with something that would once have been enormously difficult and expensive to obtain: continuous biological and behavioural information collected from people living their normal lives.

The consumer has bought the equipment and their body generates the information. Yet the greater long-term commercial value may lie in what can be learned when all those measurements are brought together.

From glucose readings to artificial intelligence

This is where the relationship becomes more complicated.

One person's glucose graph has limited research value on its own. Patterns found across very large populations can be considerably more valuable. Researchers can investigate how glucose changes alongside sleep, activity and other behaviours, whether combinations of different measurements reveal previously unnoticed patterns and whether AI can turn those relationships into personalised recommendations.

Abbott says the research will help shape future AI-powered guidance and metabolic-health innovation. With appropriate research participation and consent, information generated from people's bodies can therefore contribute to knowledge that improves algorithms and future products.

The consumer may retain rights over their personal information and have protections governing how those data can be used. That is different, however, from owning the intellectual property, algorithms or AI systems developed from research involving large populations.

The knowledge created from collective data can become part of a proprietary commercial system.

That changes the question from "Who owns my glucose reading?" to something much bigger:

"Who owns what can be learned from all of us?"

Conveniently stopping short of a medical trial

There is another unusual aspect to Abbott and Google's announcement. They describe the project as a real-world metabolic health study rather than a clinical trial. Technically that distinction is important, but commercially it is also rather convenient.

The research will combine continuous biological measurements with wearable information, laboratory results and surveys to investigate relationships between people's behaviour and metabolic health. Its findings are expected to contribute to future AI-powered health guidance and innovation. To the ordinary consumer, that may sound remarkably similar to medical research.

The important distinction is that this is observational research rather than an interventional medical-device trial. Participants aren't being assigned an experimental treatment and Abbott isn't testing an investigational glucose sensor for regulatory approval. Commercially available technologies can instead be used to observe what happens to people in the real world.

That places this type of research on a different regulatory path from certain FDA-regulated clinical investigations of medical devices.

For investigational devices, the FDA can require an investigational plan approved by an IRB, informed consent, monitoring, records and reports. Significant-risk device studies additionally require FDA approval of an Investigational Device Exemption.

The Abbott-Google study hasn't been announced as that type of investigation. Whether staying on this side of the regulatory boundary is simply a consequence of the type of research being undertaken, or also an attractive feature of how modern consumer-health research can be devised, is worth asking.

The advantages are difficult to overlook. Huge quantities of biological and behavioural information can potentially be studied outside a conventional medical-device clinical trial, while what is learned can contribute to AI health guidance and future commercial innovation. At the same time, the people generating the information have bought the consumer technology collecting it.

It is difficult not to wonder whether stopping just short of a medical trial is a feature rather than a coincidence.

Are there any safeguards?

Real-world research does not mean research without rules.

Google's Health API policy requires human-subject health research using its API to undergo independent ethical review, such as through an Institutional Review Board (IRB) or Ethics Committee. The board's stated purpose is to protect participants' rights, safety and wellbeing and it must have authority to scrutinise, modify and approve the research.

But "independent review" needs some explanation.

An external IRB can be selected by the organisation sponsoring research and commercial IRBs can be paid to carry out reviews. They are still required to exercise independent judgement, but their principal role is protecting research participants and assessing issues such as risk, consent and safeguards.

The FDA similarly describes the primary purpose of an IRB as protecting the rights, safety and welfare of human research subjects.

That is very different from independent scientific scrutiny.

An IRB does not automatically give outside scientists access to Abbott and Google's underlying dataset, independently test the companies' analysis or reproduce their findings. Nor is an IRB the same as researchers submitting their eventual results to independent medical peers for publication and peer review.

Those are significant distinctions. Ethical oversight can protect the people taking part while the companies conducting the research can still retain control of the research environment, underlying commercial data and knowledge subsequently developed from it.

An individual can therefore retain rights over their personal information while a company controls the technology, algorithms and intellectual property developed from learning across a population. Independent ethical review does not turn that proprietary research resource into one available to the wider scientific and medical community.

That distinction could become increasingly important as the datasets themselves become one of the greatest competitive advantages in health AI.

When health data become somebody else's intellectual property

Imagine that millions of people contribute different types of biological information over the coming decade.

Glucose is only one measurement. Wearables are already capable of monitoring increasingly sophisticated aspects of human physiology, while laboratory results, medical records, sleep, exercise, diet and other information can add further context.

AI can potentially identify relationships within those enormous datasets that no individual doctor, researcher or patient could discover alone. That could be extraordinarily beneficial, identifying disease earlier, revealing previously unnoticed risk factors and making genuinely personalised healthcare possible.

But who controls the resulting intelligence?

The people whose bodies generated the original information don't collectively own Google's AI or Abbott's future algorithms simply because their data contributed to research from which those technologies learned. The resulting intellectual property can remain with the organisations developing it.

That creates the possibility of a remarkable transfer of value:

Our collective biology provides the raw material, but private companies can own the machinery that learns from it.

Google has now provided a striking real-world example of how this process can work. Its insulin-resistance research used a wearable foundation model pretrained on 40 million hours of sensor data. Researchers then used wearable information alongside demographics and routine blood biomarkers to develop models capable of predicting insulin resistance.

The significance isn't simply the enormous scale of the data. Wearable users generate heart rate, activity, sleep and other physiological information about themselves. Brought together at population scale, those measurements can contribute to AI capable of extracting something new from the data: information about changes associated with insulin resistance, despite the wearable not directly measuring glucose or insulin.

The data collected yesterday for one purpose can therefore become the intelligence used tomorrow to tell us something the original sensor never directly measured.

The Nature study goes further. Researchers integrated their insulin-resistance prediction into a large language model designed to contextualise the results and provide personalised recommendations. They describe the wider approach as a scalable framework for the early detection of metabolic risk that could potentially allow intervention before progression to type 2 diabetes.

That is precisely why ownership of an individual measurement tells only part of the story. The greater long-term value may lie in the knowledge that can be extracted when millions of individual measurements are brought together, and in the AI systems subsequently developed to interpret that knowledge.

There is a scientific consequence as well as a commercial one. If the largest and richest health datasets increasingly sit within private platforms, the ability to ask questions of those datasets can also become concentrated. Universities, independent researchers, public health bodies and clinicians may not have access to anything comparable.

There are exceptions. Google has made de-identified data from its insulin-resistance study available to approved researchers, although access remains controlled rather than openly available.

The competitive advantage therefore isn't simply having a better algorithm. It can be having access to a body of human health information from which entirely new insights can be developed.

And Google's research demonstrates how that advantage can compound. A company doesn't necessarily need a new sensor measuring a new biomarker every time it wants to provide a new health insight. AI may be able to extract additional information from enormous quantities of physiological data its existing devices are already collecting.

That possibility becomes particularly significant when considered alongside the Abbott partnership. Abbott can provide direct continuous glucose information, while Google's wearable ecosystem can provide information about other aspects of physiology, sleep and activity. The planned metabolic study proposes bringing multiple forms of information together to investigate what can be learned from the relationships between them.

The long-term value of that combination may therefore extend far beyond the individual measurements being collected today.

It is not simply a question of what your wearable knows about you now, but what companies may eventually learn from the accumulated data of millions of people.

From useful health tool to health gatekeeper

Today, wearing a consumer glucose sensor is optional. People can decide whether the information is worth paying for and stop using the technology whenever they choose.

The bigger question is what happens if AI-powered health interpretation becomes substantially better than anything available without it. If proprietary systems eventually become capable of identifying health risks earlier, predicting metabolic disease or interpreting combinations of biological measurements beyond what an individual or clinician can easily understand, access to those systems could become increasingly valuable.

As these systems grow, there could also be a powerful feedback effect. Companies with the most users generate the largest pools of health data; larger datasets can improve the AI systems built from them; and better systems can attract more users, generating still more data.

What begins as competition between health apps could therefore leave a small number of platforms with an advantage that becomes increasingly difficult for competitors, public healthcare systems or independent researchers to reproduce.

At that point the relationship changes. A company is no longer simply selling a sensor. It controls part of the intelligence required to interpret what the sensor and other health technologies are telling us.

This concern is no longer confined to speculation about the distant future. A May 2026 analysis in the Journal of Medical Internet Research described consumer wearables as emerging healthcare gatekeepers, exploring how they could increasingly have the first conversation with someone about a change in their health before a doctor becomes involved.

The US provides an interesting indication of where that could lead. Wearable companies are moving beyond simply collecting health information and towards greater integration with healthcare itself. The JMIR analysis points to developments involving WHOOP and Oura, as well as Apple, Samsung, Withings and Alphabet's Verily.

The US healthcare model is very different from Britain's NHS, but the underlying issue is remarkably similar:

What happens when a commercial platform becomes the first organisation to notice a possible health problem, interpret what it means and influence what the person does next?

If a small number of companies control the largest datasets, most advanced algorithms and platforms through which those insights are delivered, they could increasingly become gatekeepers between people and knowledge about their own bodies.

What happens when AI starts predicting diabetes risk?

This question has already become considerably less hypothetical.

Just a day after the Abbott-Google partnership was announced, Google unveiled Insulin Resistance Trends, a new Health Guardian feature for qualifying Pixel Watch and Fitbit devices.

Remarkably, it doesn't measure glucose or insulin. Instead, AI analyses physiological information collected passively by the wearable over several weeks to identify changes associated with insulin resistance. Google says this can reveal changes in metabolic health without requiring a single drop of blood.

The underlying research is substantial. A Google Research study published in Nature in 2026 used wearable information and routine blood biomarkers to develop deep-learning models capable of predicting insulin resistance. Researchers described the approach as a potential framework for the early detection of metabolic risk and also integrated the prediction into a large language model designed to provide personalised explanations and recommendations. The research included 1,165 participants, with additional independent validation, and involved researchers from Google and the University of Cambridge's Institute of Metabolic Science.

But the way the resulting consumer feature is positioned is particularly interesting.

Google describes Insulin Resistance Trends as a general wellness feature. It explicitly says it isn't intended to diagnose, treat or prevent disease and, importantly, is not a prescreener for diabetes. Users should not use it to adjust medication or treatment.

At the same time, Google tells consumers that prolonged unmanaged insulin resistance can lead to deterioration in metabolic health, that identifying changes early creates an opportunity to improve it and that users can receive notifications when the system identifies something requiring their attention.

That illustrates just how blurred the boundary between wellness and medicine could become.

An algorithm can tell somebody that an important metabolic process associated with future type 2 diabetes appears to be changing while the product simultaneously remains officially positioned as neither diagnostic nor a diabetes prescreener.

And what happens next is equally interesting.

Google says users can turn to its AI-powered Google Health Coach to understand their information and receive guidance about lifestyle changes, or share the information with their doctor. Access to Health Coach requires a Google Health Premium subscription.

The pathway could therefore increasingly become:

Wearable collects data
↓
Proprietary AI identifies a health trend
↓
Commercial AI explains what it means
↓
Consumer takes action or approaches their doctor

For a UK user, that final step could mean the NHS.

Interestingly, that expansion sits in sharp contrast to the approach currently taken by the NHS. NICE recommends CGM for people with type 2 diabetes only in limited, defined circumstances, rather than routinely for the wider type 2 population. Its current quality standard identifies particular circumstances involving people using insulin, including those on multiple daily injections who cannot self-monitor using finger-prick testing because of a condition or disability and insulin-treated adults who need another person to monitor their glucose.

CGM & type 2 diabetes

Around 50,500 people with type 2 diabetes in England were estimated by NICE to be eligible for CGM in 2026/27. A small proportion of the estimated 4 million people living with type 2 diabetes in England.

Abbott, however, appears keen to court a much wider audience for glucose monitoring. Lingo takes glucose sensing beyond diabetes management and into consumer metabolic health, encouraging people without diabetes to continuously monitor how their bodies respond to food, exercise and everyday life.

That contrast is worth considering. The NHS uses CGM in type 2 diabetes in limited circumstances where there is a defined clinical rationale, while the commercial opportunity grows as more people become persuaded that continuously monitoring and interpreting their metabolic health is valuable.

For companies developing this market, widespread medical necessity may not even be required. Consumer health operates on a different threshold: people need to believe that continuously knowing more about their bodies is useful enough to pay for it. AI potentially makes that proposition considerably more powerful by turning thousands of measurements into apparently simple personalised advice.

But there is a major difference between an algorithm identifying statistical risk and a doctor making a diagnosis.

This is where the emerging gatekeeper role becomes important. A wearable capable of identifying and interpreting physiological changes could increasingly have the first conversation about a health concern before a doctor does.

In the UK, the commercial pathway may be different, but the consequence could be remarkably similar. By the time somebody contacts their GP, a commercial platform may already have told them what it thinks is happening inside their body. And with data, this is progression from the past practice of 'googling' symptoms!

A consumer receiving an AI warning about insulin resistance or other information suggesting increased metabolic risk is unlikely to treat it simply as an interesting wellness insight. For many, the next step will understandably be their GP.

And the UK already has experience of what can happen when health testing moves outside conventional clinical pathways.

Research into direct-to-consumer testing in the UK has highlighted the potential consequences of false-positive findings, including anxiety, additional investigations and increased use of healthcare services. False negatives present the opposite problem by potentially providing reassurance where further medical investigation is actually needed.

AI-generated metabolic predictions could take that problem to another scale. Unlike an occasional home blood test, a wearable continuously generates new measurements and therefore potentially new opportunities for an algorithm to identify something it considers abnormal.

Even a highly accurate system operating across millions of users could generate large numbers of people seeking HbA1c tests, GP appointments or further investigation. False positives, borderline predictions and findings of uncertain significance could add further demand.

That creates an unusual division of responsibility.

A commercial platform can generate and monetise the health insight, while the NHS may inherit the cost of confirming it, explaining it, reassuring the individual or treating whatever is eventually found.

There is also a psychological cost to consider. More information doesn't automatically mean better health. Continuous monitoring can become anxiety-provoking when people begin interpreting every change as evidence that something is wrong, particularly when an AI assigns significance to a biological trend without a clinician immediately available to explain what it actually means.

The distinction between wellness and medicine could then become increasingly difficult to defend. A product may officially provide "health insights", but if its AI tells someone that a metabolic process associated with future disease is changing, the effect on that individual can begin to look remarkably like receiving a medical warning, even when the product explicitly says that it isn't providing one.

If these platforms eventually become capable of identifying increasingly sophisticated disease risks without initial medical intervention, the important question will not simply be whether they can do it. It will be who validates those predictions, who explains uncertainty, who deals with false alarms, who carries clinical responsibility when something is missed and who pays for the healthcare that follows.

Otherwise, commercial AI could increasingly become the first to tell someone that something may be changing in their health, while conventional healthcare becomes responsible for dealing with the consequences.

Could we become dependent on the systems our data created?

This is perhaps the most uncomfortable question of all.

What begins as a consumer choosing to buy better information about their own health could ultimately contribute to a system in which society becomes increasingly dependent on commercial AI to interpret health.

Consumers provide the biological data from which those systems can learn, yet they do not collectively own the resulting AI and may eventually have to pay to access its most useful conclusions. Independent researchers and public healthcare systems may not have access to the underlying data required to build anything comparable.

Healthcare providers could also become increasingly reliant on commercial platforms if those systems demonstrate that they can identify patterns, predict risks or personalise treatment more effectively than existing methods.

The danger isn't that an AI suddenly "owns" somebody's body or medical records. It is much more subtle:

Dependency.

Once a technology becomes sufficiently useful, opting out can become increasingly difficult even when participation technically remains voluntary. We have already seen versions of this elsewhere in digital life, where a handful of platforms have become difficult to avoid because communication, commerce or information increasingly flows through them.

The US experience deserves attention here. The JMIR analysis asks what happens when wearable platforms move beyond measurement and towards influencing the route into healthcare itself.

Britain has a very different healthcare structure, but that does not necessarily make it immune. A commercial platform does not need to own a GP surgery or hospital to influence what happens next. It may only need to become sufficiently trusted that consumers arrive at their NHS appointment already believing its interpretation of their health.

Health could therefore become the next frontier for the platform economy. And health is fundamentally different from social media, search or online shopping because the consequences of being outside the system could ultimately affect how well somebody understands their own body.

CGM is becoming much bigger than diabetes

There is another reason this development deserves attention from people already using CGMs.

Continuous glucose monitoring was developed as medical technology for people with diabetes. In England alone, more than 380,000 people received FreeStyle Libre on NHS prescription in 2024/25, according to NHS prescribing data.

For people with type 1 diabetes, an important distinction is that CGM isn't simply providing interesting information about health. It is a medical tool used every day to make treatment decisions, including decisions about insulin. That is fundamentally different from using glucose information for general wellness or metabolic-health insights in people without diabetes.

Lingo takes the same fundamental ability to continuously measure glucose into consumer wellness. Interestingly, Abbott increasingly reframes the language around the technology too. Rather than foregrounding CGM, a term closely associated with diabetes and medical treatment, Lingo is described as a “biowearable”, or BGM, while Abbott talks more broadly about blood glucose monitoring and metabolic health. The underlying measurement may be familiar to anyone who uses a CGM, but the language makes it sound like something quite different: everyday health technology for a much wider audience.

Google then adds another layer by combining those readings with a much wider ecosystem of health information and applying artificial intelligence to interpret them.

The result starts to look less like a glucose sensor and more like part of a personalised health platform, with a potential market considerably larger than diabetes alone.

Abbott isn't alone in moving in this direction. Dexcom is also building a much broader health platform, expanding beyond traditional diabetes management through Stelo and its wider interest in consumer and metabolic health.

The biggest CGM manufacturers increasingly appear to see a future extending far beyond simply supplying sensors to people with diabetes.

The long-term prize could be the platforms that interpret continuous human health data.

Editor's comment: who will own the future of personalised health?

There is enormous potential for good in this technology. Continuous health information combined with powerful AI could identify risks earlier, make healthcare more personalised and potentially prevent disease.

The concern isn't that we should stop that innovation. It is who controls what comes afterwards.

We use the sensors, our bodies generate the data, companies control the datasets, those datasets are used to build the intelligence, companies own the systems that interpret it and we increasingly depend upon those companies to tell us what our own bodies are saying. The resulting algorithms and AI can then become commercial intellectual property controlled by the companies developing them.

At the same time, the underlying data resource isn't generally available to the wider scientific community that might otherwise use it to advance medical knowledge independently.

If proprietary systems subsequently become the best way of interpreting our health, the same population whose data helped make them possible could become increasingly dependent on accessing them.

There is another possible cost too. Commercial platforms may increasingly generate health risks and predictions outside conventional healthcare, while doctors and public health systems are left to investigate the results, manage false positives, reassure anxious users and ultimately provide treatment.

The issue becomes still more complicated when the company interpreting our health is also in a position to influence what we do next. Consumer technology has spent decades becoming exceptionally good at directing attention and behaviour. Applying those same techniques to decisions about our health deserves considerably greater scrutiny.

That is a very different relationship from buying a glucose sensor. It raises questions about whether health intelligence developed from enormous populations should become concentrated inside a small number of commercial platforms, what meaningful alternatives will remain available and whether people will retain genuine choice if proprietary AI becomes increasingly embedded within healthcare.

The Abbott-Google partnership is only one development in a rapidly changing industry and it may ultimately produce genuinely useful health technology. But it also gives us a glimpse of a possible future in which we use the sensors, our bodies generate the data, companies control the datasets, those datasets are used to build the intelligence, companies own the systems that interpret it and we increasingly depend upon those companies to tell us what our own bodies are saying.

Much of this is presented as progress: greater convenience, better health insights, earlier warnings and improved safety. Who would argue against any of those things? But that narrative also makes it remarkably difficult to question where medtech is taking us, who ultimately controls the systems being created and what we may be giving up along the way.

That is why the biggest question raised by this partnership isn't simply who owns your glucose data.

It is who will control the knowledge created from it, who will influence what we do with that knowledge, and who will deal with the consequences when it starts telling us we might be ill.

External sources

Abbott: Abbott and Google launch first-of-its-kind partnership to transform everyday health through glucose insights and AI

Google: Google Health and Abbott partnership

Fierce Biotech: Abbott teams with Google Health on AI-powered health insights

Google: Health API User Data and Health Research Policy

FDA: Investigational Device Exemption (IDE)

BMJ: research on direct-to-consumer self-testing, false positives and healthcare use

FDA: Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices

Nature: Insulin resistance prediction from wearables and routine blood biomarkers

Google: Health Guardian / Insulin Resistance Trends

JMIR: Meet the New Health Care Gatekeeper: Your Wearable

NICE: current type 2 diabetes/CGM guidance