The EU seeks the formula to regulate ChatGPT

New technological systems like ChatGPT They have become popular just when the artificial intelligence law is being negotiated in Brussels and the European Union he wants to take advantage of the moment to find the formula to regulate the social, political and ethical doubts that they raise.

Initially, when the European Commission proposed the regulations in April 2021, it was not intended for this, but in the long and tortuous path of the European legislative process, these artificial intelligence models have exploded and all the institutions of the EU they now believe that they must be addressed.

While discussions continue in Brussels, Italy has for the moment banned the use of ChatGPT, considering that Open AI -the company that created and continues to refine this tool- does not comply with the European Data Protection Regulation (GDPR).

The Italian authorities accuse the company of a lack of transparency regarding the collection of users’ personal and banking data and of lacking a system to verify the age of those who use ChatGPT, despite the fact that it is aimed at people over 13 years of age. and they have given him until April 30 to adapt to the law.

The Spanish Agency for Data Protection also opened this Thursday an investigation into the company for a “possible breach” of the GDPR and the European Committee for Data Protection created a working group to exchange information with all the authorities of the EU, Iceland, Liechtenstein and Norway that have doubts about it.

Green MEP Marcel Kolaja is convinced that Open AI “has interest” in resolving these issues but considers that these models must continue to be regulated, according to what he told EFE.

Besides, “some of the concerns that these systems can raise are not related to privacy”explains to EFE the liberal MEP Dragos Tudorache, one of the speakers for the artificial intelligence law in the European Parliament.

Among them, he mentions the possibility that these models capable of simulating human conversations -which are also developed by other companies such as Google- give answers based on information protected by copyright, the risk of being used to spread false news or the influence they can have on vulnerable people.

And in this sense, he recalls the recent suicide of a young Belgian in his thirties after having talked intensively for six weeks with the Eliza chatbot of the American company Chai about his concern about climate change.

As currently drafted, the artificial intelligence law prohibits real-time mass surveillance systems in public spaces -except to prevent crimes and locate possible victims- and vetoes models that use subliminal techniques to “substantially” alter ” the behavior of a person, without being aware of it.

In addition, it classifies as high risk a whole series of artificial intelligence systems with very specific uses that can only be introduced to the market if they respect the fundamental rights and values ​​of the EU.

For example, those used by companies to select candidates for a job, those used by financial institutions to assess the solvency and establish a person’s credit rating or those used to guarantee the security of the water supply. , gas and electricity.

But now, “many of us agree that we cannot afford to have nothing in the text” of the law to regulate systems like ChatGPT, says Tudorache.

Both he and Social Democrat Brando Benifei, also a rapporteur for the regulations, want developers to be transparent about the way they train their models, to have human supervision and to prohibit the spread of illegal content.

Negotiations between political groups to set the criteria will begin this Monday and neither Benifei nor Tudorache dare to say how they will conclude, since they admit divisions between parties.

EPP MEP Axel Voss bets first on “understand” how ChatGPT works and ask Open AI for explanations about its reliability, according to EFE. And in any case, he is in favor of the self-regulation of the sector rather than writing new laws that can take years to be approved.

“In three years, we may have totally different problems. You never know if ChatGPT will still exist,” she points out.

On the other hand, the Community Executive also recognizes “the need to consider specific standards” for this type of tool, as explained to EFE by sources from the cabinet of the Commissioner for the Internal Market, Thierry Breton, and the same is the opinion of the Council of the EU.

However, none of the MEPs interviewed is in favor of pausing the training of these systems for six months, as has been proposed by more than 2,000 experts, academics and businessmen, in order to have time to develop protocols that guarantee the security of artificial intelligence.

“It is not feasible that we stop the development and rather, it is important to accelerate a global effort to regulate it correctly”says Benifei, who believes that we must join forces in the United Nations, the G20 or the OECD.

Source: EFE