This French startup is using blockchain to train medical AI

To train the large language models empowering AI, companies need data – the more structured and richer the data, the better. That’s why provenance matters when it comes to AI training data, which brings numerous challenges.

Take healthcare. Training AI for medical purposes requires access to medical records at some level. Those medical records contain personal details of the most private nature you can think of. This is the kind of data that regulations, such as GDPR, care so much about they have their own categories for them.

Then there’s the issue of ransomware: data this valuable is bound to attract cyberattackers. And when it’s held by institutions that have a duty of care, such as hospitals, they become even more attractive to criminals.

At CES, we spoke to a French company that aims to keep all the above in check. Galeon is using blockchain technology to keep data anonymised while training AI medical models, and has also helped hospitals remain operative during cyberattacks. And while the topic didn’t come up during our interview, it has also created the equivalent of an NFT for organ donors, an idea designed to help ensure those who want to donate their organs to science have better control over it.

Useful applications for blockchain and NFTs? No wonder Galeon won three Innovation Award Honoree awards at CES 2024! Interview below, edited for clarity.

Recommended reading: Real NHS doctor asks AI for medical advice

If you could, please start by introducing yourself and telling me about the company.

I’m Dr Loïc Brotons, a medical doctor in anesthesiology and intensive care. I’m the CEO of Galleon, a healthcare company. We provide electronic health record management to 13 hospitals in France, where we are based.

What kind of solutions do you offer?

Our approach is to create highly structured medical data, the kind needed to train medical AI with a high degree of quality. We do so by leveraging blockchain, which helps train AI in a decentralised manner. That particular technology is called Blockchain Swarm Learning. The data remains locally stored in each hospital and it’s the learning algorithm that moves from one hospital to another, tapping into that data to be trained.

Or, in other words, an algorithm can be trained train without actually having to move data around for it to access it. As an example, let’s say that a startup wants to train an algorithm to detect lung cancer. To do so, the first requirement is access to 5,000 X-rays with a potential diagnosis of lung cancer. Those X-rays might be available across ten different hospitals. With our approach, the X-ray data never leaves the hospital yet the algorithm is still trained on it all. The end result for the startup is they get an algorithm that is trained on the data it needs to be trained on, and done in a secure manner that doesn’t expose patient data to any risk.

Are your typical clients mostly hospitals?

Correct, all our customers are hospitals. But while we are the heart and core of hospitals, ultimately our product impacts doctors and patients.

You seem to have won a few awards this year…

We’ve won one Innovation Award for the blockchain part of our technology and two Innovation Awards for the cybersecurity part. That’s because our system enables hospitals to keep running even during a cyberattack.

We already had two hospitals that saw a cyberattack destroy their local systems, but Galeon switched to safe mode which kept everything running during the cyberattack in a way that meant the hospital didn’t have to move the patients. Usually, when there’s a cyberattack, everything is shut down and patients get moved to another hospital. It’s a huge problem. For example, we just had a German journalist tell us that Frankfurt Hospital was cyberattacked in September and it’s still not open again.