The big tech firms rallying together against the EU’s AI Act, which they argue is “over-regulating” powerful foundation models like GPT-4, also known as General Purpose AI, will almost certainly be disbanded. These companies lobbying against the type of strict regulation contained within the AI Act are in danger of negating what the Act set out to achieve.
Throwing out the proposal, Axel Voss, Member of the European Parliament for the Cologne/Bonn region (CDU), said the regression of the Act on Foundation Models in favour of such minimal “mandatory self-regulation” would require the same insufficient benchmark for standards on issues like “transparency, cybersecurity and information obligations”, standards that are well-defined in the Act.
He suggested on social media the proposal was backwards, non-sensical and would dilute necessary strong regulations to manage all AI technologies, including the advanced models of General Purpose AI which power the famous generative chatbot, ChatGPT.
He also said: “The regulation of #AI has a global dimension, even the US executive order calls for it. We cannot fall behind. Of course, I have always said and maintain that we should not over-regulate. But we cannot simply ignore or downplay the risks that come with it, including on foundation models”.
The attempt to derail the AI Act which is entering its final phase to be implemented came on Nov. 23 when a letter was written to the EU Commission trying to protest against “harsh” regulations. The letter was undersigned by 33 companies, allegedly containing supporting data that only 8% of EU companies use AI. The commotion of Chat GPT and OpenAI is only evidence that AI innovation is happening worldwide, but it’s innovation, carrying significant harms too, needs to be carefully managed.
MEPs also assembled a close ring defending stricter regulation for those powerful AI models. After initial consensus on a tiered-approach, the mediation with techs derailed even more seriously when German, French and Italian governments seemed to take the oppositions’ side, pushing back on obliging to the Act’s stance on foundation models. This recommenced the chaos within Parliament.
Axel Voss has equally shared his view that the Parliament has bowed to AI Act protesters in this way with a final Act looking to be diluted.
He argued for what impact no AI Act would actually have on companies:
“Such outcome of the AI Act negotiation would dump all the regulatory burden and compliance cost to the companies downstream: those that train foundation models to develop AI applications for specific use cases. These are mainly SMEs. Frankly, I do not understand this”, he wrote. It makes the compliance requirements for innovative companies even harder to meet.
“It makes Europe even more dependant on a handful of, mostly foreign, dominant players.
Isn’t it better, even for the EU’s aspirant AI champions like #Mistral and #AlephAlpha, to work with the regulators and outcompete #OpenAI and the rest with products that are designed with inbuilt compliance?” he continued.
Companies are just played the “narrative of BigTech against the DMA” in wanting to be similar European wannabe AI foundation models rather than seizing the opportunity to join forces with the regulators.
