In June this year, the European Parliament voted to pass the world's first "Artificial Intelligence Bill", and now the bill is facing a critical moment, with negotiators trying to finalize the final details this week. According to the legislative agenda, the European Parliament will conduct "tripartite" negotiations on the bill with the European Commission and member states. The negotiated bill will take effect on the 20th day from the date of publication in the EU Official Journal, and will be applicable 24 months after it takes effect. A more optimistic guess is that it will be applicable in 2026.
It would be the world's first comprehensive bill to regulate artificial intelligence and would further cement the EU's position as a global trendsetter in controlling the tech industry.
But at the last minute, ongoing disputes over how to regulate underlying models such as ChatGPT have stalled the process. Big tech companies are lobbying against excessive regulations they say stifle innovation, while European lawmakers want to add protections for the cutting-edge artificial intelligence systems the companies are developing.
Meanwhile, global alliances including the United States, Britain and other G7 nations have joined the race to develop protections for the technology, with researchers and human rights groups warning that generative AI poses an existential danger to humanity and risks to daily life.
Nick Reiners, a technology policy analyst at political risk consulting firm Eurasia Group, said: "The EU Artificial Intelligence Bill will hardly become the global gold standard for industry regulation. It may not be agreed before the European Parliament elections next year. There are too many things that need to be finalized in the final round of negotiations on Wednesday. Even if they work late into the night as expected, they may have to be postponed until next year."
When the European Commission published the above-mentioned artificial intelligence draft in 2021, it almost did not mention general artificial intelligence systems such as chatbots. Instead, AI systems are divided into four risk levels (from minimal to unacceptable), which essentially serves as product safety legislation.
But that’s all changing with the boom in generative artificial intelligence, which is incredibly powerful at composing music, creating images, and writing articles. However, generative AI has also raised concerns that the technology could be used to launch large-scale cyber attacks or create new biological weapons.
These risks prompted EU lawmakers to extend their regulations to the base model. Also known as large language models, these systems give chatbots like ChatGPT the ability to create new things, which is very different from traditional artificial intelligence, which processes data and uses predetermined rules to complete tasks.
Altman regained control of OpenAI after the company experienced dramatic infighting last month. In this regard, the European Commission's director of industrial policy, Titieri Breton, said that at least now the situation is clear, companies like OpenAI are defending their own business rather than the public interest.
At present, the main opposition forces within the EU against the artificial intelligence bill are France, Germany and Italy. The three largest economies in the EU tend to adopt a self-regulatory approach similar to the US model. The change is believed to be an effort to help local generative AI companies such as French startup MistralAI and Germany's AlephAlpha.
Additionally, EU negotiators still need to resolve a number of other contentious issues, including a proposal to ban real-time public facial recognition entirely. Some countries want exemptions so law enforcement can use it to find missing children or terrorists, but rights groups worry that would effectively create a legal basis for surveillance.
The EU's three government ministries will have one of their last chances to strike a deal on Wednesday. Even after they reach a deal, the EU's 705 lawmakers must still approve the final version. The vote needs to take place before April next year in order to take effect before the European Parliament elections next June. If they are unable to do so in time, the legislation will be delayed until later next year, when new EU leadership may take office with a different view on AI.
In a panel discussion last week, Romanian MP Dragos Tudorache, who co-led the negotiations on the artificial intelligence bill in the European Parliament, said: "It is very likely that this is indeed the last good opportunity. We still need more time to negotiate. In short, the outlook is still unclear."