The World Economic Forum’s MINDS programme has recognized FLock.io for its privacy-preserving artificial intelligence work with two National Health Service trusts, a project aimed at improving eye disease detection and diabetes management without moving sensitive patient records outside secure NHS systems.
The recognition places the company’s federated learning platform among a group of real-world AI deployments being highlighted for use in highly regulated sectors. The NHS-related work involves Moorfields Eye Hospital and University College London Hospitals, where AI models are being trained across local clinical data sources while patient information remains within protected health networks.
The approach is designed to solve one of the biggest barriers to healthcare AI: how to train useful models on large and diverse medical datasets without centralizing confidential records in one location. In conventional AI development, data is often gathered into cloud-based environments for training. In healthcare, that can raise legal, ethical, cybersecurity and public trust concerns. Federated learning takes a different route by allowing models to learn from data where it already sits.
FLock.io’s inclusion in the WEF programme comes as hospitals, governments and financial institutions are looking for AI systems that can deliver practical benefits while complying with strict data governance rules. The company says its system combines federated learning with blockchain-based verification to help validate model updates, protect data sovereignty and reduce exposure to manipulation.
How the NHS projects are using the technology
At Moorfields Eye Hospital, the technology is being used to support AI training for eye disease detection through imaging data. Eye care has been one of the clearest early use cases for medical AI because diagnosis often depends on detailed scans, images and patterns that machine learning systems can be trained to recognize.
The goal is not to replace clinicians, but to support faster, more consistent detection of disease in settings where demand for specialist care is rising. Eye hospitals and ophthalmology departments face growing pressure from ageing populations, diabetes-related complications and backlogs in screening services. AI tools that can safely learn from medical images without transferring patient records may help hospitals improve triage and identify cases that need urgent attention.
At University College London Hospitals, researchers from University College London are testing AI-based glucose monitoring alerts using data from more than 400 patients. The system is intended to improve diabetes management by predicting glucose changes and warning users before levels move into dangerous ranges.
That work is expected to move into a broader multi-continental glucose prediction trial this summer. The trial is set to involve 100 patients across the UK, Europe, the United States and China, giving researchers a chance to test whether the system can perform across different populations, clinical settings and data environments.
The international nature of the planned trial is important because AI systems can struggle when they are trained in one setting and used in another. A model built primarily on data from one hospital, one country or one demographic group may not perform as well elsewhere. Federated learning is meant to reduce that weakness by enabling multiple sites to contribute to model training while keeping data under local control.
Why federated learning matters in healthcare
Federated learning allows AI models to be trained across decentralized data sources. Each hospital or participating site processes its own data locally and sends only encrypted model updates for aggregation. The underlying patient records do not need to leave the institution.
In simple terms, the model travels to the data rather than the data travelling to the model. That distinction is central to the technology’s appeal in healthcare, where patient privacy rules can make large-scale data sharing difficult or impossible.
The method can also help address data inequality in AI. Smaller hospitals, regional health systems and institutions in different countries may hold valuable clinical data but lack the ability or permission to send it to a central AI developer. Federated learning creates a structure for collaboration without requiring every participant to give up control of its records.
For the NHS, this issue is especially sensitive. The health service holds vast quantities of patient information, and previous attempts to expand data use have faced public scrutiny. Any AI system used in this environment needs to show that it can protect privacy, preserve trust and comply with regulation while still producing clinically useful results.
FLock.io says its platform supports deployment on local hardware, giving participating institutions more control over where computation happens. That may appeal to public-sector health systems that want to avoid sending sensitive workloads into external cloud environments without clear safeguards.
Potential savings from diabetes prevention
FLock.io estimates that AI-driven diabetes prevention could reduce NHS expenditure by more than £100 million a year if the technology cuts current diabetes management costs by just one percent. Diabetes care is estimated to cost the NHS more than £10 billion annually, including treatment, monitoring, hospital admissions and complications such as kidney disease, nerve damage and vision loss.
The estimate highlights why diabetes is a major focus for AI developers and health systems. Even small improvements in prevention, early intervention or day-to-day management can produce large savings because the condition affects millions of people and requires continuous care.
The company says around 14,000 users already use its technology through diabetes management applications in the UK and parts of Asia. While that figure suggests early adoption, broader clinical impact will depend on trial results, regulatory review, integration with care pathways and acceptance by clinicians and patients.
Glucose prediction tools must also prove that they are reliable in everyday conditions. Diabetes management is affected by meals, exercise, medication, sleep, stress and individual metabolism. A useful AI alert system must balance sensitivity with accuracy, warning users early enough to act without creating unnecessary alarm or alert fatigue.
Blockchain verification and model integrity
FLock.io’s platform combines federated learning with blockchain verification. The company says this helps record and validate model updates while reducing the risk that faulty or malicious inputs could influence training.
In federated systems, one challenge is that many parties may contribute updates to a shared model. If a participant sends poor-quality, biased or manipulated updates, the combined model could suffer. Verification tools are intended to make the training process more transparent and accountable.
The company reports that its architecture has delivered a 37% boost in model accuracy, a 44% reduction in ownership cost, a 63% faster deployment rate and 80% lower energy use per update. These figures are presented by the company as evidence that decentralized AI training can be both privacy-preserving and operationally efficient.
Such claims will likely be assessed closely by healthcare institutions and regulators. In clinical environments, performance improvements must be measured against patient safety, bias, explainability and real-world effectiveness. A model that performs well in a technical benchmark still has to be validated in clinical practice before it can influence care decisions at scale.
Energy use is another important consideration. AI systems can require significant computing resources, especially when trained on large datasets. If federated systems can reduce unnecessary data movement and run efficiently on local infrastructure, they may offer environmental and cost advantages compared with some centralized approaches.
Wider use in regulated industries
The technology is not limited to healthcare. Federated learning is also being explored in banking, insurance, government services and other sectors where data cannot easily be pooled in one place.
Banks, for example, may want to collaborate on fraud detection without sharing customer records. Public agencies may need AI tools that work across departments without breaching data protection rules. Manufacturers may want to improve predictive maintenance models across factories while keeping operational data confidential.
These industries face a similar problem: valuable data is fragmented across organizations, jurisdictions and systems, but legal and commercial restrictions prevent easy sharing. Federated learning offers a possible path around that obstacle, though implementation remains complex.
Technical issues include data quality, network reliability, cybersecurity, model governance and the risk of hidden bias. Organizational issues can be just as difficult. Participating institutions must agree on standards, responsibilities, evaluation methods and liability if a model makes a harmful recommendation.
That is why recognition through programmes such as WEF MINDS matters. It signals that privacy-preserving AI is moving from a research topic into practical pilots involving major institutions.
Sarawak pilot expands the cross-border picture
Outside the NHS partnerships, the government of Sarawak in Malaysia is conducting a sovereign AI pilot with FLock.io. The initiative includes healthcare projects that the company says could be scaled across hospitals in the United States, Europe and China.
The Sarawak project is part of a wider global interest in sovereign AI, a term used to describe systems that allow governments or institutions to develop and control AI capabilities within their own legal, cultural and data environments. For many countries, sovereign AI is linked to economic policy, national security and digital independence.
In healthcare, sovereign AI has an additional dimension. Medical data is among the most sensitive information a government can hold. Cross-border collaboration can improve research and model performance, but it also raises questions about consent, jurisdiction, storage, accountability and oversight.
A federated approach may help bridge those concerns by allowing hospitals in different countries to train shared models without exporting raw patient records. If successful, the model could provide a template for collaboration between Asia-Pacific and European institutions, particularly in areas where disease patterns, treatment access and clinical resources vary widely.
WEF MINDS recognition
The WEF’s MINDS programme includes organizations working to expand real-world applications of AI in regulated and industrial environments. FLock.io is being recognized alongside companies such as Lenovo, Occidental, TCL Industries, Hisense Hitachi and KUKA, with Accenture involved as a programme partner.
The inclusion of companies from manufacturing, technology, industrial systems and healthcare-adjacent fields reflects the broader direction of AI deployment. The most important next phase may not be consumer chatbots or experimental tools, but systems that can operate safely inside hospitals, factories, energy networks and public infrastructure.
For AI companies, regulated sectors offer large opportunities but also impose high standards. A system that handles medical records, financial transactions or industrial controls must be secure, auditable and reliable. The cost of failure can be far higher than in consumer software.
FLock.io’s position in the programme suggests growing interest in distributed machine learning as a way to meet those standards. The company has argued that its design reduces data leakage risks while allowing large organizations to collaborate on model development across borders.
Digital asset market context
FLock.io also operates within the broader market for decentralized AI and open computing networks. Public financial records cited by the company show that the underlying token network had generated more than $2.7 million in direct user fees by late last year. The startup has also raised $11 million from venture capital firms to expand its open computing system.
The wider market for AI-linked digital assets has grown quickly and has recently been valued above $20 billion. These networks often aim to reward users for contributing computing power, data access, model training or verification services.
For traders, the sector has become a closely watched part of the digital asset market. However, clinical partnerships and token activity should be viewed as separate issues. A healthcare pilot may demonstrate technical adoption, but it does not automatically determine the value or long-term viability of a related digital asset.
Useful measures of real activity can include public ledger data, wallet growth, active node counts, user fees and the number of completed computing tasks. These indicators can help distinguish networks with practical usage from projects driven mainly by speculation. Even so, digital asset markets remain volatile, and token prices can move for reasons unrelated to healthcare progress or technology performance.
What comes next
The next major milestone for the healthcare work is the planned glucose prediction trial involving patients across multiple regions. Its results will be closely watched because diabetes management is a high-cost, high-impact area where better prediction tools could make a measurable difference.
For Moorfields, progress in AI-supported eye disease detection could also strengthen the case for privacy-preserving model training in imaging-heavy specialties. Ophthalmology, radiology, dermatology and pathology all rely heavily on images, making them natural areas for AI development.
The broader question is whether federated learning can move from promising pilots to routine use in hospitals. That will require not only technical performance, but also trust from clinicians, patients, regulators and health system leaders.
If FLock.io and its partners can show that AI models can be trained across institutions without compromising privacy, the approach could become an important part of healthcare’s digital infrastructure. The recognition from the WEF’s MINDS programme gives the company visibility, but the decisive test will come from clinical outcomes, regulatory acceptance and the ability to scale safely across real-world health systems.
For deeper insight into privacy-preserving finance and data, explore our guide on data tokenization and why it’s important.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
