“Is this what life in Europe will look like in the age of artificial intelligence?”, asked a few days ago the technology director of Hugging Face, a platform that allows developers and researchers to train and test generative algorithms. It accompanied the reflection with a meme that shows several SpongeBob characters dancing in the street while another watches them from a dark room. A joke with a clear and real message: the technological giants of Silicon Valley are prohibiting Europeans from playing with their new artificial intelligences. They don’t like the rules that the EU has imposed in this park and the veto is their way of showing that the ball is theirs.
In recent days it has been confirmed that three long-awaited major systems will not be available in Europe. The first is the Llama 3.2 model from Meta, owner of Instagram, WhatsApp and Facebook. It is a multimodal AI, which means it can process and generate information from text, images, documents and graphics. It is the first major system with an open approach, which means that unlike others like ChatGPT, it allows you to download and customize it according to the needs of each person or organization. The developer community believes that this will be a big step forward in the integration of AI tools.
The other two models censored by tech companies in the EU are the voice version of ChatGPT and Apple Intelligence. Both are based on OpenAI technology and aspire to be what virtual assistants like Siri or Alexa promised: systems capable of carrying on a conversation with humans. Able to detect their emotions, their needs or the tone of the dialogue.
The three companies claimed that the reason they did not deploy these tools in the EU was the new regulations affecting their businesses. “The unpredictable nature of the European regulatory environment,” according to a Meta spokesperson. He is the only one who had an obvious conflict with the European authorities. This happened when the social media company tried to make it harder for Europeans to opt out of llama training with their photos and comments on Facebook and Instagram. After being reprimanded by data protection authorities, Meta decided not to use this information to train its AI, but Llama would be banned in the EU.
“There are growing concerns that this could lead to ‘two-speed AI’: more advanced AI for the rest of the world and less capable AI for the EU,” said Phillip Schmid of Hugging Face, who is also an American platform. “While regulations are intended to protect, they can also hinder innovation. We must therefore quickly find a balance to ensure that progress does not stagnate,” he continued: “European Commission, we would be happy to provide our support!
This is the position they expressed in a letter signed by entrepreneurs and researchers with expertise in AI, including Mark Zuckerberg and Daniel Ek, founder of Spotify. “Fragmented regulation means the EU risks missing out on the AI era,” they threaten. One option, they explain, is to make a “modern interpretation” of the laws, “so that AI innovation happens here at the same scale and speed as elsewhere.” The other is to “continue to reject progress, to contradict the ambitions of the single market and to watch the rest of the world rely on technologies to which Europeans will not have access”.
The EU can continue to reject progress and watch the rest of the world rely on technologies that Europeans will not have access to.
Letter from entrepreneurs and researchers to the EU
Pressure even comes from companies that have deployed their AI tools in Europe, such as Google. Kent Walker, its president of global affairs and general counsel, the executive who sits down to negotiate with policymakers, is currently in Europe to talk about regulation.
“My journey comes at an important time, where we are seeing notable advances in AI tools,” he wrote on LinkedIn on Wednesday. “However, as science advances, much of the public debate revolves around risks, from AI safety to concerns about discrimination and abuse,” the official said.
I fear that we don’t talk enough about the biggest risk: that of missing out on the next technological revolution.
Kent Walker
— President of Global Affairs and General Counsel at Google
For Walker, the opacity with which AI is developed or the possibility that it reflects the prejudices of its creators, such as machismo, racism or classism, are secondary factors. “Addressing these issues is important, but we also need to recognize that long-term innovation generally leads to more safety, not less. And I fear that we are not talking enough about the biggest risk: that of missing out on the next technological revolution,” he stressed.
“We only ask for transparency”
Brussels did not react to this attack. Official Executive sources said they do not comment on “individual decisions”, but reiterated that “all companies are invited to offer their services in Europe, as long as they respect European legislation”. “We all know that AI offers many benefits, but it also comes with risks. We need AI that people can trust. This is why the European AI law has established clear rules to safeguard the fundamental rights of citizens,” they emphasize.
Other sources, however, are clear on the diagnosis. “They’re trying to apply pressure, like big tech companies have always tried. The objective now is the new European Commission,” warns Carme Artigas, former Secretary of State in charge of Digitalization and Artificial Intelligence.
Artigas led negotiations on the artificial intelligence regulation during Spain’s EU presidency and now co-leads the UN advisory committee on AI. He points out that technology giants have already been exerting this pressure for years to relax EU regulations on this technology. “It’s a push for regulation. The same that happened in California when the governor vetoed a law approved by a Parliament,” he recalls.
In the United States, state governors have the ability to veto bills approved by the representative houses. This Sunday, Gavin Newsom, governor of California (home of Silicon Valley), used this power to reject a key regulation imposing security requirements for the development of advanced AI models. The law aimed to ensure safety before deploying high-risk AI systems, but Newsom argued it was too broad and could stifle innovation in the state’s tech industry.
Transparency gives European companies the assurance that the system is not a black box, that it does not violate copyright, that it is not discriminatory. This is why the adoption of an AI that does not have a clear business model being delayed by six months does not delay us at all.
Carmen Artigas
— former Secretary of State for Digitalization and Artificial Intelligence.
Artigas also rejects tech companies’ arguments against regulation. “It is not true that there is regulatory fragmentation within the EU. The two regulations on AI. “as the Digital Markets Act and the Digital Services Act are directly applicable, i.e. they are exactly the same rules for all EU countries”, answers the expert , who is also part of the advisory board of the consulting firm Llorente y Cuenca. .
“We are only asking for transparency. None of these companies can give up in a market of 600 million consumers because transparency is required, because there cannot be mass adoption if people do not trust”, Artigas elaborates: “Transparency gives companies Europeans the assurance that the system is not a black system. box, it does not use copyrighted material without permission, it is not discriminatory. This is why the adoption of AI that does not have a clear business model is delayed by six months, which does not delay us at all.”
In Ursula Von der Leyen’s new European Commission, powers relating to large technology companies are distributed among several commissioners. On the one hand, Teresa Ribera will assume the functions of Competition, the team responsible for preventing abuse of dominant position. The mission of promoting technologies developed on the continent will fall to the Finnish Henna Virkkunen (PPE), while the tasks of regulating the internal market, including the digital market, will fall to the Frenchman Stéphane Séjourné (Liberals).
