AI ACT: how to analyse the new European legal framework?
Alexandre Rispal - The Council of Europe[1] has just ratified the text on responsible IAs, voted through the European Parliament last March 2024. This legal framework will also be open to non-European countries. The AI ACT is a milestone. Faced with the rise of these technologies, Europe is adopting legislation[2] that aims to promote responsible AI practices with systems "(...) safe, transparent, traceable, non-discriminatory and environmentally friendly". We use algorithms every day to get around, to shop, to book our doctor's appointments. The AI ACT is unique in the world. This regulation is in line with the fundamental rights of citizens and CSR[3] in the same way as the GDPR on respect for personal data. AI ACT aims to protect with responsible AI models. And this with a clear framework.
AI ACT and CSR, is it a major differentiator for Europe?
AR - Accused of slowness, bureaucracy and a wait-and-see attitude in the face of the Gafams and Internet champions who have managed to impose their models, Europe is fighting back. It wants to make its mark on the market for cutting-edge innovations and stand out from the crowd. And how? By promoting responsible AI that are more ethical, more secure, and less subject to the risks of manipulation for citizens. The approach is intended to be pragmatic and more human. And it's a clear differentiator compared with other players, such as the American hyperscalers[5] and their generic AI systems, which consume huge amounts of data and energy. It is also a differentiating factor in that, as in China, we will not see the emergence of AIs using biometrics to identify citizens in real time, which could restrict freedom of movement, for example. The risks are identified in the text of the AI ACT.[6]
How not to lose the battle to the American hyperscalers?
AR - This is precisely THE trap to avoid. We don't want to be like the American hyperscalers. But defend our vision of innovation without being left behind from a technological standpoint. After all, we already missed out on the Internet revolution 30 years ago in the face of the Gafams. Not a single European champion has managed to emerge due to a lack of clusters and structured funding for fundamental research. And without fundamental research[7], there can be no cutting-edge innovation. The aim of AI ACT is certainly to protect, but also to unleash the creative energies of our research laboratories. And why not bring back our engineers and mathematicians who have left for Silicon Valley for lack of anything better in France. Especially as the AI ecosystem matures. Mistral AI[8] the French startup has developed a new language model in just 1 year. This entrepreneurial dazzle - 30 employees today - would not have been possible without a mature ecosystem. Europe does not want to miss the AI trend seen as a technology that will profoundly transform our societies. AI ACT is intended as a response to the frantic race [9] for ever more AI. Innovating means capitalising on a technologically reliable and secure model while drawing on our specific cultural characteristics. Above all, innovating means not copying and pasting the hyperscaler model.
Responsible AI, what business benefits in the insurance sector?
AR - I've been observing insurance and its technological developments for 20 years. Artificial intelligence is putting the sector in turmoil. Learning models, like the progress made in Machine Learning, Deep Learning, have enabled us to considerably improve the intelligent analysis of documents. The collection of supporting documents (housing certificate, bank account details, ID) is now done almost in real time. This saves time for customers. It's also a productivity booster for businesses that have to deal with multiple customer relations challenges. And it works if you intelligently combine these AIs in the intelligent document processing as during the appraisal of customer files. This is what ITESOFT's Streamline Business solution offers, for example, which incorporates artificial intelligence to facilitate underwriting or claims management, while at the same time securing the customer journey. AI serving the needs of the business perfectly illustrates the desire of the European regulator: reliability, pragmatism and risk-free for the user.
Footnotes:
[1] Council of Europe adopts first international treaty on artificial intelligence, press release 17 May 2024.
[2] AI ACT to come into force from 2026.
[3] Principle of social and environmental responsibility (CSR).
[4] The 2024 European elections will take place from 6 to 9 June. Every 5 years, the citizens of European Union countries elect their representatives to sit in the European Parliament in Strasbourg.
[5] Google, Microsoft, IBM, Amazon, Meta.
[6] Unacceptable risks : cognitive-behavioural manipulation, social scoring, real-time and remote biometric identification system. High risks: AI systems linked to product safety: toys, aviation, cars, medical devices, lifts. Limited risks : the consumer must be informed that they are interacting with an AI and the generated content must mention that it was generated by an AI.
[7] AI, our ambition for France. Anne Bouverot, Philippe Aghion. Odile Jacob Editions (2024)
[8] AI Générative : Mistral AI or the open source route with a commercial licence " - LeMagIT, 25 July 2023.
[9] In which countries are we betting the most on artificial intelligence? Data from Stanford University's "2023 AI Index Report". The United States comes out well ahead with $248.9 billion, followed by China ($95.1 billion) and the United Kingdom ($18.2 billion). In eighth place, France follows closely behind Germany with investments of $6.6 billion over this 20213-2022 period.
Source: Regulating facial recognition in the EU. European Parliamentary Research Service (2021).
Source : Future of Life Lethal Autonomous Weapons Pledge - Future of Life Institute.
