How can we better manage artificial intelligence, fight against bias and abuse, and work towards the implementation of a trusted AI based on transparency, without hindering innovation? To answer these questions, the Think Tank Digital New Deal presented its report “Trusted AI, a strategic opportunity for digital and industrial sovereignty”. The authors have developed 18 proposals to better manage AI and promote an “AI of the Enlightenment”.

We asked Julien Chiaroni, Director of the Grand Défi en Intelligence Artificielle at the Secrétariat Général pour l’Investissement, and Arno Pons, from the Think Tank Digital New Deal, five questions.

In a few words, can you introduce the Digital New Deal think tank and its role in the French and European digital landscape?

The role of our think tank is to participate in the creation of a European and humanistic “Digital Age of Enlightenment” by enlightening public and private decision-makers on the challenges of digital sovereignty.

We are fully dedicated to issues related to the centralization of power and the capture of value by “Big Tech”. We provide concrete solutions through regulation and cooperation. Our publication and advocacy activities have a considerable impact: for example, the 2.4 billion fine against Google Shoppings, the GAFA tax, etc.

What do you call the “do-tank”?

Our Do-Tank activity is the natural extension of the think tank.
It aims to create a “3rd digital way”, an alternative to the American and Chinese offers, by supporting economic sectors and local authorities in the creation of trusted ecosystems. The Do-Tank sets up technological alliances, such as the creation of a “data space” within the framework of Gaia-X.

How do you set up an AI label?

Trust labeling is a key to the trust ecosystem, as it allows manufacturers to certify their compliance with the law. It is systemic in the sense that the opposite (private self-labeling) amplifies the opacity of monopolistic companies.

It is a chain of trust, with bodies at each stage, which must respect a “separation of powers” between each: values in charters, regulations translated into harmonized standards, labels of complementary audit, certification and investigation mechanisms. Our recommendation is therefore to give preference to official/state labelling and to federate the ecosystems to create a European NIST or LNE, capable of issuing labels for the EU with an extraterritorial vocation.

What are the principles and values that you attach to the “AI of the Enlightenment”?

The challenge is to prevent algorithmic biases from becoming economic biases (American AI that is purely commercial and serves the interests of Silicon Valley oligopolies) or political biases (Chinese AI that monitors and rates citizens in the name of the interests of a totalitarian power).

The AI of the Enlightenment is, on the contrary, a European AI, ethical, transparent, which respects our humanist values. The general public and economic actors must indeed have “confidence” in these AIs, which govern their lives (access to consumer credit, job recommendations, health …).

It is a question of sovereignty, but also of competitiveness, especially for our industries, which can use trusted AI as an opportunity to (re)conquer markets hitherto dominated by Big Tech.

What are the strategic axes presented in the report that will help strengthen trust in AI?

To strengthen trust in AI, we need to create the conditions for trust through regulation and create an ecosystem of trust through cooperation. The common thread running through these recommendations is the issue of governance.

We propose to create bodies that do not yet exist: a common AI-Data governance and an industrial AI alliance (as exists for Data with Gaia-X).

One of the strong points of our report is the proposal to create an InfraTech (bringing together innovative software infrastructure startups). The goal is to create an intermediate layer that will lower the cost of access to the market (“level playing field”) and to contain and above all conquer the Big Tech market (which benefits from the “winner takes all” effect).

InfraTech is a potential lever of scalability for FrenchTech and European industry if we succeed in creating this end-to-end package that will allow the multitude to become a credible offer in the face of the GAFAM monopolies.

Translated from 5 questions à Julien Chiaroni et Arno Pons sur le rapport sur l’IA de confiance de Digital New Deal