European Union (EU) legislators reached a historic deal on Friday over negotiations on an artificial intelligence (AI) regulation bill dubbed the EU AI Act, which was first passed in June, the European Parliament announced in a statement.
According to the statement, the regulation aims to “ensure that fundamental rights, democracy, the rule of law and environmental sustainability are protected from high-risk AI while boosting innovation and making Europe a leader in the field.”
A post on X by Thierry Breton, the EU’s Commissioner for the Internal Market, announcing the deal was quickly dunked on by proponents of unregulated acceleration of AI technology, who view the bill as a form of onerous regulation that will hold back innovation in the sector.
“The #AIAct is much more than a rulebook — it’s a launchpad for EU startups and researchers to lead the global AI race,” Breton wrote. “The best is yet to come! 👍”
“Thank you for ensuring Europe is slow and last to fail at the AI revolution,” wrote @mysticaltech. “You did not learn from the cookie banner debacle did you.”
EU regulations require that websites ask users for consent to store cookies, which are small data tracking files that are stored on a user’s machine to identify themselves to websites they’ve visited before. The EU regulation is responsible for many websites requiring pop-ups asking for cookie consent before accessing their pages and has come under frequent criticism for breaking website functionality, not actually changing data retention practices or protecting privacy, and just being plain annoying.
“This is the way the future ends,” @SwiftOnSecurity wrote, riffing on the same theme. “Not with an explosion of progress but a cookie warning.”
What is the EU AI Act?
The Financial Times describes the new law as laying out a “restrictive regime” for the emerging technology, and much of the criticism online has reflected fears and disdain about that being the outcome of the law, hampering innovation, all at the same time without any real protections being secured by the legislation.
The law defines three “unacceptable risk” areas where AI systems are considered a “threat to people” and will be banned. Those include “cognitive behavioural manipulation of people or specific vulnerable groups;” social scoring, i.e. “classifying people based on behaviour, socio-economic status or personal characteristics;” and “real-time and remote biometric identification systems, such as facial recognition.”
However, the law carves out exceptions in these unacceptable risk areas in certain cases with court approval, such as biometric identification systems after a “significant delay” to, for instance, prosecute serious crimes.
The law also defines high-risk areas where AI systems will be required to be registered in an EU database. That includes when it involves products that the EU regulates under safety legislation, like toys, aviation, cars, medical devices, and elevators, as well as eight specific high-risk areas. Those areas are “biometric identification and categorisation of natural persons,” “management and operation of critical infrastructure,” “education and vocational training,” “employment, worker management and access to self-employment,” “access to and enjoyment of essential private services and public services and benefits,” “law enforcement,” “migration, asylum and border control management,” and “assistance in legal interpretation and application of the law.”
The law also requires generative AI tools like ChatGPT to provide transparency labels disclosing AI-generated content, to design their models so that they can’t generate illegal content, and to publish summaries of the copyrighted data they use to train their models on.
On X, the EU statement was also mocked because of its prediction that the EU AI Act would serve as a “launchpad” for innovation in the industry.
“It’s a launchpad alright, it launches tech workers directly to the United States, very cool, nice job @ThierryBreton!!” joked @JObeseya.
“Being European used to be cool…. used to be sexy!” posted @hazelmulhare. “Now this is what we are known for.”
While the reaction to Breton’s tweet was overwhelmingly negative, not everybody was gloomy about the Act.
“there were a lot of dunks and derision about how EU is a stifling bureaucracy with no tech sector,” wrote @n00rdung, “silly decrepit fools which conveniently ignores that, on the very same day, a certain french startup just dropped a magnet link to their open source LLM.”
That was a reference to Mistral, a Paris-based AI startup which released a language model today which can be run locally on a personal computer without a connection to the internet and performs roughly on par with Chat-GPT3.5
“bureaucracy and regulations like this are pretty much a feature of europe at this point,” @n00rdung went on. “those regulations are at this point a form of protectionism; local lawyers navigate through, foreigners in theory give up,” pushing a positive tone that the laws may help EU AI innovators compete.