2024-04-19 14:17:08
The Artificial Intelligence Commission to “contribute to making France a country at the forefront of the AI revolution” submitted its report on March 13. It makes 25 recommendations so that France can benefit from the AI technological revolution. Research in the health sector is targeted, with a major challenge: health data. The question of the balance between protection of personal data and freedom for the benefit of innovation once once more comes to the heart of the debate. A look back at this long-awaited report, the results of which we are now awaiting.
Analysis by Alexandre Fievée, associate lawyer, and Alice Robert, Lawyer Counsel of the firm Derriennic Associés for La Veille Acteurs de Santé.
The Artificial Intelligence Commission for “ help make France a country at the forefront of the AI revolution » was installed in September 2023 by the Government. Its report is based on a widely shared observation that has now become commonplace, namely that AI is a major challenge for the future of Europe and France. He points out that AI is an unavoidable technological revolution that affects absolutely all areas of activity. He adopts a reasonable scenario on the expected impact, emphasizing that AI should not arouse either excessive pessimism or excessive optimism (“ We do not anticipate mass unemployment or automatic acceleration of growth »).
However, the commission believes that France, like Europe, must take up the challenge of AI, “ otherwise we will not have control of our future “. A challenge that France, like Europe, can meet, the Commission believing that we have the assets to do so. But there is no time to lose.
To achieve this, the report identifies six main lines of action:
Immediately launch a national awareness and training plan; Structurally reorient savings towards innovation and create, in the short term, a “France & IA” fund of 10 billion euros; Make France a major hub for computing power; Facilitate access to data; Assume the principle of an “AI exception” in public research; and Promoting global AI governance.
The development of AI in health requires the release of data
The report clearly identifies the Health as one of the key future areas for the development of AI.
First observation: AI makes it possible to understand a considerable volume of available data, which human intelligence cannot process. “ More than 5 million scientific articles are published each year, half of which are in the field of medical research alone.indicates the Commission. It is obviously impossible for a researcher or a team of researchers, even a high-level one, to be able to read them, let alone evaluate and analyze them. »
Second observation: data is an essential ingredient for recent developments in artificial intelligence. And if this data is not necessarily personal, it is clear that many of them are of a personal nature. “ Exploiting the potential of artificial intelligence and enabling its deployment in the service of humans therefore requires that researchers, developers and innovators have access to massive, reliable, easily manipulated data whose representativeness and quality can be assessedunderlines the Commission.
In a context of rapid technological development and increased competition, this access must also be able to be opened to them quickly and the data used without excessive constraints, at the risk of further favoring existing players or seeing others take over. appropriate our research and innovations, by getting ahead of them in their experimentation and dissemination. »
Third observation: access to data is often complicated and the constraints are considered excessive by AI stakeholders, whoever they may be (companies, researchers, laboratories, public and private institutions, associations).
Regulatory constraints to be lifted for access to data
According to the Commission, the constraints are of two orders. First of all, certain French rules and practices are more restrictive than the European framework when it comes to the processing of personal data. If the GDPR has, with the principles of freedom and responsibility, completely reversed the logic of the law which prevailed in France since the “Informatique et Libertés” law of January 6, 1978 (in application of which the processing of personal data was based on prior authorization or declaration procedures to the CNIL), the constraints are still too strong in the health sector.
« There are still prior authorization procedures not provided for by law European », regrets the Commission. “ This is particularly the case for access to health data for research. A simplified procedure for declaring conformity to reference methodologies exists but it is far from being generalized. In practice, the simplified procedure remains the exception compared to the prior authorization procedure because the slightest deviation from these methodologies involves going through a prior authorization which can involve up to three levels of prior authorization. »
Then, the Commission notes “ a growing gap between logic centered on the protection of the individual and the evolution of modes of collective use of data “. According to the Commission, several key notions of the GDPR are poorly adapted to the functioning of AI:
the notion of “data controller”, “ for which the distribution of responsibilities between the developer who has trained a generative AI and who makes it available to third parties and the end user of the system for his own needs does not necessarily appear self-evident » ;
the notion of “purpose of processing”“which conditions the nature of the data that can legally be used and to which the consent of the persons concerned relates is also more complex to understand, given the numerous possible uses of a generative AI once it has been trained » ; the very notion of “personal data”, “ which constitutes the key to applying the GDPR, raises questions in a growing context of use of collective data ».
Even the anonymization of personal data, which allows one to “exit” from the personal data protection regime of the GDPR, does not seem suitable because “ technology increasingly opens up possibilities for re-identification of anonymized data ».
Towards a reform of the “Informatique et Libertés” law?
The Commission recommends “ to eliminate prior authorization procedures for access to health data and to reduce response times from the CNIL “. She adds that this development should be accompanied by a reform of the mandate given to the CNIL, to include a “ innovation objective “. She ends by suggesting the idea of a “ collective governance » of the data which might pose “ the milestones of an evolution of the legal framework which would better take into consideration the evolution of the methods of use of data. » It now remains to know the plan that the government wishes to put in place on the basis of this report…
Read the other analyzes by Alexandre Fievée and Alice Robert
1713587079
#Health #freeing #data #facilitate #innovation #research #Analysis
