Brussels warns against ‘paranoia’ when regulating generative AI

Receive free Artificial intelligence updates

We’ll send you a myFT Daily Digest email rounding up the latest Artificial intelligence news every morning.

One of the EU’s most senior officials has warned against being “paranoid” or too restrictive when regulating generative artificial intelligence because fear of the technology would stifle innovation.

Věra Jourová, the European Commission’s vice-president for values and transparency, told the Financial Times that the bloc’s impending legislation should not be based on “dystopian” concerns.

“There should not be paranoia in assessing the risks of AI. It always has to be a solid analysis of the possible risks,” said Jourová, one of two commissioners overseeing the enactment of the EU’s landmark AI law.

“We should not mark as high risk things which do not seem to be high risk at the moment. There should be a dynamic process where, when we see technologies being used in a risky way we are able to add them to the list of high risk later on.”

When asked, she agreed that too much regulation posed a threat to technological and business innovation.

The EU has been at the forefront of the race to regulate AI but others, including the US and China, are debating their own controls on its development and use. The UK is hosting a global summit on regulating AI next month.

Jourová’s comments come as the commission, European parliament and member states start the final stretch of negotiations to finalise the AI act, two-and-a-half years after the commission proposed the legislation. Officials hope to conclude the discussions by the end of the year.

The negotiations come after businesses raised concerns that the use of generative AI could lead to the manipulation of public opinion through the use of deep fakes. There are also worries among some members of the European parliament about the technology’s ability to create original content that could violate copyright laws.

The commission’s original draft focused on the defence of human rights, such as privacy, and was based on regulating the risks of the technology.

But the parliament has introduced legal obligations on the creators of so-called foundation models and their applications, such as ChatGPT, to make them liable for how the technology is used.

Developers of generative AI models such as ChatGPT would be required to reveal AI-generated content and provide summaries of copyrighted data employed in their training to fight disinformation and illegal content.

Brando Benifei, one of the parliament’s leading negotiators, said: “A principles-based approach doesn’t seem enough to have binding rules and actionable rights for citizens, businesses and institutions.”

Businesses have also warned of the risks that regulating specific technologies have on innovation. The Computer and Communications Industry Association, a US-based trade body which represents developers of AI, said last week that “applying stringent requirements to the basic building blocks, or cutting-edge technology as such, simply would create legislation that cannot stand the test of time”.

“The same applies to banned practices, which need to be clearly targeted at specific cases of unacceptable use of technology,” it added.

Jourová, who is overseeing the passage of the legislation with internal market commissioner Thierry Breton, said she would make her case at a G7 forum on AI in Kyoto later this week for the EU’s model to be the leading example on how to regulate the technology.

“We see this as setting the standards for the democratic world but China is not part of this because in the world of technologies we are more rivals than partners,” she said, adding that there was a “low level of trust” between Beijing and the group of advanced economies.

Jose Fernandez, US under-secretary for economic growth, energy and the environment, told journalists in Brussels on Monday that AI was a key area of collaboration between the US and EU as the trading partners try to counter the influence of China in the digital realm.

“The pitfalls of artificial intelligence in some countries are questionable, [and] something that concerns us,” he said. However, there were also “huge benefits” to the technology that the US and EU “want to make sure we maximise”, he added.