About European AI regulation race

03 January 2024 | Julien Vandanjon-Rancoule | Category (en-gb)

Market Watch

Tempus Fugit

While the press refers to 2023 as "the year of AI", bids are running high on technology startups involving artificial intelligence, and almost all governments are capitalizing on their most promising colts by knowingly ignoring all consumer protection laws, or timidly and superficially addressing any prospect of serious regulation of the subject.

In June 2023 European Parliament introduced the AI Act policy draft , after a two years processing, totally rewrote after public release by the end of 2022 of OpenAI GPT3 revealing the gamechanging power of Generative AI. European lawmakers are expecting to be the first group of nations in the world to offer a regulatory canvas to build on, and thus hoping to trigger industrial adaptations towards a common "manageable" standard. Now early 2024 internal dissensions already slowed down the process and some voices already claims it may be too late to harness the beast. What does it mean ?

It may sound contradictory but the challenge here is to be the first to - metaphorically - "build brakes" to negociate a seat in the A-train. Even if there is high level engineers and wealth in western europe to support the research in Generative AI and Multimodal Large Language Models, which are the core of future B2C applications, the business lanscape and linguistic patchwork of Europe compress it so it stays deeply entangled with internal competition between countries such as France, Germany, Spain, etc.

These struggles blocks any option for the emergence of a real "European Champion" able to compete with the GAFAM (Google, Amazon, Facebook (-Meta), Microsoft and their asian challengers from Alibaba, Tencent and so on.

It also means that while shy European experiments and natio-centric university laboratories are trying to catch up, the front-runners are widening the gap in relative impunity, where the boundary between "finished" commercial products and "full-scale experimentation" is becoming increasingly blurred.

See : France24 TV Channel Analysis after AI Act first adoption following 37 hours debates in December 2023

For the winners, the benefits are so massive as to make any form of financial coercion negligible

Almost all Generative AI actors in 2023-2024 market opted for the TokenStock/Rental model. As half the planet wants to experiment with AI - and basically can do without needing an extra hardware investment - the Generative AI subscriptions skyrocketted with a limited need for marketing budget. Everything you need is a basic app, a couple of instagram or tiktok influencers, a youtube video and you're in.

Since nobody really understands how it works, and there is no regulation for products stamped "AI something" you could basically sell your old dishwasher by branding it AI and nobody will really raise an eyebrow. The US Federal Trade Commission is trying to warn consumers about AI Fraud in software applications only using good old algorithms and there is no doubt Europe will capitalize on the AI Act to, at least, establish a vocabulary for fraudulent AI marketing even if the range of its applications is not fully discovered.

Everybody knows the main actors built their supremacy on massively nurturing their initial LLMs with some copyrighted data and possibly by training them to replicate authors and artists styles which will never really be able to claim proportional financial retribution for it. The Generative AI rocket went so far so quick that it is almost impossible to sue them for unauthorized derivative works seriously, so the only way to reduce the gap with more ethical alternatives is by boosting the influx of combined efforts both from financial and editorial asset owners joint-forces.

See ISACA Paper on Understanding AI Act

Up to 255 Millions Users for AI Tools in 2023, not counting casual mobile app and social media users

Source : Statista

Sample : Midjourney generation of famous Ghost in the Shell Character Motoko Kusanagi by Masamune Shirow in Jamie Hewlett style

So... Is the battle worthless ?

Yes... and no. For those who missed the train and now dream of regulations as levers to put on hold an already thriving international market to catch up, it sure is a candid dream. Even if earlier in 2023 Italy managed to put AI wave on hold for a while by using consumer protection existing guardrails, they couldn't hold the blocus for long considering the industrial and competitive pressure. On another hand, by agitating a regulatory sandbox and framework under the nose of the actual leaders such as Google, Microsoft/OpenAI, Adobe, Amazon etc. the European Politics certainly hope to put this as a deal : "If you don't comply with our guidelines you will force us to turn to local or eastern providers to fulfill our future uses of AI in critical industries such as healthcare, intelligence, automotive, defense, etc." Thus inducing a huge loss of markets and benefits for western providers, and challenging the (already) weakening grasp on the "old continent".

From this perspective, we can see a glitch of the highly strategic and political challenge opened by the quick spreading of Generative AI. Rarely an innovation went so quickly on the front stage of international balance of powers and the players of this game began almost 60 years ago is only beginning to reveal its challengers. From our point of view Madrid 2023 was not a gamechanger, at least it allowed some of the european countries to showcase some of their contestants. But from a global sight, the cautious but democratic stances taken by Europeans are a good barometer of the game ahead.