It takes a village to train AI, but the data stays at home

The security, data protection, and regulation of artificial intelligence solutions have become critical issues worldwide. At the same time, organizations are seeking more effective ways to leverage their data. The solution developed by Turku UAS’ SA-FlaaS project is part of this global growth story.

Let’s imagine a situation in which a manufacturing company would like to use its error reports to train an AI model. The model would be significantly more useful if the training data also included error reports from a competing company that manufactures the same product.

Using shared training data, the AI could learn to identify how different work stages, processes, and work cultures contribute to errors. Both companies could then develop more efficient operating models, improve workplace safety, and proactively prevent problems. Everyone would benefit from the new insights generated by AI.

However, the problem is obvious: organizations and companies are often reluctant to share their data. In healthcare or government operations, it isn’t even a matter of willingness. Legal and strictly regulated barriers prevent the sharing of patient data.

The SA-FlaaS (Secure and Accurate Federated Learning as a Service) project developed a solution to this problem. The project built a platform based on distributed machine learning that allows AI models to be trained without the need to transfer or disclose sensitive data to third parties.

This is the result of long-term research and development efforts. In recent years, researchers at Turku UAS have achieved international success in the development of federated learning technologies. An innovation related to brain cancer research also ranked among the top entries in an AI competition sponsored by Intel. This international recognition is proof that SA-FlaaS’s AI solution is gaining ground in a highly competitive and extremely demanding market.

Federated learning has commercial potential

The solution developed by the SA-FlaaS project, funded by Business Finland, is based on federated learning. Each organization retains its own data within its own systems and contributes to training the model without sharing data with other parties.

This innovation, driven by healthcare needs, continues the work of the INSIGHT research group’s other R&D projects. The project portfolio has since expanded with even more ambitious goals.

SA-FlaaS has recognized the commercial potential of this solution. If an AI solution developed for healthcare can be applied to any industry, it represents a significant opportunity for scaling.

A secure AI solution is attracting interest

Not all RDI projects end up in the conference rooms of Japan’s Ministry of Health, Labour and Welfare (MHLW). However, representatives from SA-FlaaS managed to secure an audience with the ministry, and a cooperation agreement was signed with the prestigious Hiroshima University. Another significant Japanese partner is the Osaka Heavy Ion Therapy Center.

A tour conducted in the United States also attests to the international interest. Heikki Lassila, the project’s Business Champion, attended numerous investor and partnership meetings in the spring of 2026. Invitations came from, among others, Cleveland Clinic Innovations, Nvidia, and UC Davis Health. These discussions provided valuable insight into the needs of U.S. stakeholders.

“With our expansion into Asia, we can conduct much better international research. Our data sets are growing, but patient information doesn’t cross national borders,” says Teemu Herttua, a senior advisor who has just begun a researcher exchange program at Hiroshima University.

The security of AI solutions is currently one of the most critical issues in the field. As the use of AI expands at an accelerating pace across various industries, so do the demands for data protection, transparency, and reliability.

Solutions such as SA-FlaaS may form the foundation for how organizations build secure AI services across organizational and national boundaries.

SA-FlaaS at a glance

SA-FlaaS (Secure and Accurate Federated Learning as a Service) is an AI platform developed by Turku UAS that enables the training of machine learning models without the need to transfer sensitive data between organizations.

• Enables the training of AI models without transferring data
• Particularly well-suited to the needs of the healthcare sector and other data-intensive industries
• Based on internationally recognized innovations in federated learning
• A project funded by Business Finland (August 1, 2024 – July 31, 2026)

Learn more about the project

Contact