The rapid development of new technologies makes it increasingly difficult to tell whether content was created by a human or a machine. Italy has just banned the use of ChatGPT on the grounds that it does not comply with European data protection regulations, but the risks of this and other new generative technologies could be even more sophisticated. Brussels wants to ensure there is as little confusion as possible, and to that end recalls that it aims, among other things, for all content generated by artificial intelligence (AI) to be mandatorily duly warned.
More information
“Anything generated by artificial intelligence, be it text – everyone knows ChatGPT – or images, there will be an obligation to report that it was created by artificial intelligence,” Internal Market Commissioner Thierry Breton said on Monday. on the French radio station Franceinfo. The senior European official recalled that almost exactly two years ago, the European executive tabled a proposal for a regulation on artificial intelligence that “provides AI developers, implementers and users with clear requirements and obligations in relation to the use specifically for AI”, as the Commission itself summarizes its proposal.
Breton, who has stressed that if adopted, Europeans will be “the first” to propose legislation of this type, has expressed his hope that the text will be voted and adopted in the European Parliament this month, an essential condition regulations must be pushed forward and adopted by the twenty-seven as quickly as possible. According to parliamentary sources, the European Parliament has on its agenda the presentation of the plan and its vote in plenary in the last week of April. If this schedule is respected, the goal of the French Commissioner, who is responsible, among other things, for increasing European technological sovereignty and applying the rules regulating this vast area, and who wants the regulation to come into force in 2025, could be achieved.
Rapid advances in AI have long worried Europe. When Brussels presented the proposed regulation in April 2021, the tools based on generative AI were indeed able to create text, images or music from a series of instructions that – for now – have reached their maximum expression with ChatGPT.
The proposed European regulation – legislation that is directly applied in the 27 countries without having to be transposed into national regulations – provides for four types of “risks”: First comes the “unacceptable” and therefore prohibited , Category This includes uses , which, like in China, enable what is called social scoring, the social scoring system that determines a person’s credibility or reputation based on various factors, including data or activity on the Internet. This is followed by “high risk”, where technologies such as the CV scanning tool used to assess and rank job applicants or some medical applications are received, which must be subject to certain legal requirements. A third category would be that of “AI with specific transparency obligations”, which includes impersonation bots like ChatGPT (although without explicit mention, which didn’t exist yet) and finally those with “minimal or nonexistent”. is allowed without restriction.
Breton’s reminder of the need to clearly warn that content has been generated by an AI applies to both the ‘high’ risk category and the third, which calls for specific obligations (both categories are not mutually exclusive, the Commission points out) . In particular, the technologies that fall into these categories must inform users that they are interacting with an AI system unless there is something obvious and when emotional recognition systems or biometric categorization are applied to them. In addition, so-called deep fakes (content that imitates a person’s voice or appearance) must have a label that informs them about it.
The latest example of how easy it is to fool half the world (or more) with AI-generated content happened this weekend when a photo of Pope Francis in a white Balenciaga puffer coat turned out to be deep fake with the generative artificial intelligence tool Midjourney. Its creator is a 31-year-old construction worker from Chicago who, as he told BuzzFeed, “just wanted to have fun making psychedelic art.”
Shortly before that, last week, the alarm went out from the producers of this type of technology and content: more than a thousand high-level businessmen, intellectuals and researchers related to generative AI signed an open letter demanding a six-month moratorium on the development of this technology to reconsider its consequences. Among the signers is Elon Musk, owner of Twitter but also founder of OpenAI, the company that developed ChatGPT.
you can follow THE COUNTRY Technology on Facebook and Twitter or sign up here to receive our weekly newsletter.