TheThe French government is proposing the deployment of a platform named Health Data Hub (HDH) to develop artificial intelligence applied to health. The HDH aims to become a one-stop access to all health data.
The data concerned are those of hospitals, pharmacies, shared medical records and research data from various registers. The amount of data hosted is likely to explode, especially with the emergence of genomics, imaging and connected objects. It is expected that this data will be stored at Microsoft Azure, the public cloud of the US giant Microsoft. This choice is at the center of our concerns.
Gafam (Google, Apple, Facebook, Amazon and Microsoft), start-ups and even insurers could access health data and the financial power they represent, if these companies show that their research projects can have a use for the “public interest”, a relatively vague concept.
In addition, the use of Microsoft is framed by paid licenses. Although discussions are being held to ensure the reversibility of the US platform, it seems difficult to change. We know the risks of digital captivity, including contracts between Microsoft and hospitals.
A breach of medical confidentiality?
The US government has adopted in 2018 a text called Cloud Act, which allows US justice to have access to data stored in third countries. The president of the National Commission for Informatics and Liberties (CNIL) told the National Assembly in September that this text is contrary to the General Regulations on Data Protection (GDPR), which protects European citizens. Concretely, patients could be subject to a breach of medical confidentiality, which constitutes a personal as well as symbolic danger, the integrity of the oath of Hippocrates being questioned.
In addition, the HDH is developing on a centralized model, resulting in a higher impact in the event of hacking. One might think that Gafam offers ultra-secure solutions. This argument does not hold. Indeed, attacks often come from within, that is to say, personnel having access to the data.
Although the data hosted by the HDH is de-identified, complete anonymity is impossible because it is sufficient to cross a limited number of data to re-identify a patient. In addition, the medico-administrative database of the National System of Health Data (NSDS), integrated into the HDH, has been criticized by the CNIL for the obsolescence of its encryption algorithm.