Cold shower for Emmanuel Macron. While the Head of State and the various governments since 2017 pride themselves on having positioned France well to become a world leader in artificial intelligence, parliamentarians who are experts on the subject are clearly not of this opinion.
In the United States, artificial intelligence elevated to the rank of tsar
Requested by the offices of the National Assembly and the Senate to take stock of the new developments of this technology in the context of the generative AI revolution, the Parliamentary Office for the Evaluation of Scientific and Technological Choices (OPECST) , co-chaired by an LR senator, Stéphane Piednoir, and a deputy from the presidential camp, Pierre Henriet (Horizons, former Renaissance), presented this Wednesday, December 4, a comprehensive 334-page report.
Written by three parliamentarians – senator LR Patrick Chaize, PS senator Corinne Narassiguin and RN deputy Alexandre Sabatou – and voted unanimously by the members of OPECST, this document provides a very mixed assessment of the French strategy on AI at work since 2018. It thus establishes 18 recommendations to rectify the situation, with the hope – probably in vain due to the current political chaos and the disaster of public finances – that they will be translated “ quickly ” in ” operational measures ».
Among them, five relate specifically to the AI Action Summit. The event will be held in France, in Paris on February 10 and 11. While welcoming the organizers' desire not only to focus on the existential and security risks of AI, the current program appears ” incomplete » to the authors. The latter believe that two dimensions considered priorities, education and digital sovereignty, have been forgotten.
The Elysée wants to put “general interest” at the heart of the Paris Summit on AI
A small arm strategy
To qualify the French artificial intelligence strategy announced in 2018 and completed in 2022, which mobilizes 1.9 billion euros in total, the authors of the report bring out the bazooka.
« Rather than announcing a strategy without objectives, without governance and without monitoring tools aimed at “making France a world leader in AI”, it is appropriate to put in place a real public AI policy, with objectives, real means including governance worthy of the name, and monitoring and evaluation tools », defy the parliamentarians. And to call for “ abandon of thethe policy of the start-up nation with its armed wing French Tech, as elitist as it is inappropriate ».
French Tech Tremplin, a buoy in the sea for those forgotten by entrepreneurship
The OPECST is thus a continuation – with added freedom of tone – of a recent report from the Court of Auditors. The magistrates estimated in April 2023 that the priority given to research in the French strategy had made it possible to “ avoid a dropout “. But they also underlined the insufficiency of its resources, its management and the monitoring of actions.
The Interministerial Committee on AI, in its March 2024 report, went even further. According to him, the State urgently needed to invest “ at least 5 billion euros per year to prevent France from falling into historic decline », as its co-president, Anne Bouverot, explained to The Tribune.
« Unfortunately, the report of the Interministerial Committee on AI had no effect in the context current policy and budget », deplores Senator Corinne Narassiguin (PS). “ The repeated vacancies in the position of national strategy coordinator since 2018, and its lack of real authority, reveal a poorly defined function. The strategy remains unmanned and evolves according to the President's announcements and communication strategy. », adds MP Alexandre Sabatou (RN).
“Investing in artificial intelligence means investing in growth” (Anne Bouverot and Philippe Aghion)
Technological autonomy
To rectify the situation, 13 of the 18 OPECST recommendations focus on improving the French strategy. Without quantifying its cost, the authors reiterate the requirement for a massive investment. In the name of the “ national sovereignty » and “ strategic autonomy » faced with the domination of the United States and China, they indicate that France and Europe must “ develop an autonomous sector across the entire artificial intelligence value chain ».
Translation: from energy to applications, including semiconductors, infrastructures and language models. “ Europe is a leader in only one segment of the AI value chain, the very precise segment of chip engraving, thanks to ASML” is written in the report. If European companies “remain in the race in model designs, AI applications and services”, their technological dependence on essentially American players in terms of processors (Nvidia), and computing and cloud infrastructures (Amazon, Microsoft, Google) is a “ sovereignty problem with major geopolitical implications », According to Senator Patrick Chaize (LR).
On the governance side, the report calls for the AI strategy to be managed “ at the highest level ”, with an attachment of the coordinator “ to the Prime Minister ”, rather than “ a very small team attached to the General Directorate of Enterprises in Bercy “. They also want OPECST to be tasked with monitoring and evaluating the national strategy.
The parliamentarians also suggest the creation of “ regional entertainment centers », in close collaboration with universities, research centers such as Inria and companies, to mobilize and lead the French AI ecosystem, and thus promote collaborations, particularly with the private sector.
This is why they are asking for the renewal of the Confiance.ai program, stopped in 2024 despite its reduced cost of 45 million euros over the period 2021-2024. This allowed manufacturers to collaborate with academic players within the framework of a secure and sovereign platform, to integrate trusted AI systems into their processes. Groups like Airbus, Atos, Renault, Safran, Thales and Air Liquide had benefited from this to transform more quickly thanks to AI.
This extremely ambitious report submitted to Macron to make France a world champion of artificial intelligence
Training, AI Grenelle and protection of creation
Parliamentarians also insist on training for schoolchildren, students, workers and even the general public, based on the Finnish model. The latter is held up as an example for its “continuing training programs” intended for all.
The objective: to demystify AI to promote its deployment in society and in the world of work. All while remaining vigilant, through regular studies, on “ the impact of AI on employment, the social fabric (including inequalities) and cognitive structures » especially children.
“We must promote a scientifically informed and rather optimistic vision of AI (…). General AI [qui surpasse les humains pour toutes les tâches cognitives, Ndlr] remains an unlikely prospect for the moment and the question of existential risks is not a priority,” they write.
The peaceful adoption of AI by society can only be achieved through social dialogue, they add. Hence the recommendation to integrate AI into corporate social dialogue, and to organize discussions on a national scale, such as “ AI Grenelle ».
Finally, parliamentarians do not avoid the cultural issues of generative AI. Firstly, by being in charge at the global level on the creation of standards, particularly ethical and environmental, for AI. Then by ensuring sovereign control of data from French culture, and creating datasets around French-speaking cultures for training AI, as Spain does to preserve its culture in the face of the American steamroller.
Simplifying global governance
With this in mind, the five recommendations of the report concerning the Summit for Action on Artificial Intelligence on February 10-11, essentially aim to obtain concrete progress for simplified governance of AI at the global level.
The parliamentarians want France to “ lays the groundwork » of global AI governance « under the aegis of a single international organization “. This would be integrated into the UN, rather than the ten or so parallel projects currently moving forward. “ It is necessary to create a specialized institution member of the United Nations, whose skills would extend from the international coordination of the regulation of AI to the fight against the global digital divide », Points out the report.
As for the European level, parliamentarians hope that the Summit will lead to the launch of a “ European AI cooperation program », which would include at least France, Germany, the Netherlands, Italy and Spain. He would be responsible for defining a “ European AI pathway » going further than supporting innovation and regulating the AI Act.
