The European Parliament adopted on Thursday unprecedented regulations in the world on artificial intelligence. Among the validated measures: the possibility of filing a complaint against AI (Artificial Intelligence) systems.
The European Parliament’s communication teams define the event as ” one step closer to the first rules on artificial intelligence “, an almost modest position in the face of what nevertheless resembles a victory, no doubt out of pure awareness of a still long road on the road to regulation. MEPs adopted, on Thursday, May 11, the world’s first rules on artificial intelligence, which will require language models such as ChatGPT to comply with certain requirements.
Faced with generative AI, the European Union calls for total transparency
These rules are brought together under what is called the “AI Act”, a 144-page long text that will contribute to the regulation of the sector within the European Union. 84 MEPs voted for, 7 against and 12 abstentions. Elected officials have insisted, through the amendments, on ensuring that AI systems have human control. The values of safety, transparency, traceability, non-discrimination and respect for the environment are in the spotlight.
In particular, MEPs have inserted into the text obligations for all companies that provide reference AI models. They will have to guarantee to the European Union a high level of protection of fundamental rights, but also of health, security, the environment, democracy and the rule of law. They will all have to register in the EU database.
Generative AIs like GPT, from Open AI, will be subject to additional transparency requirements. For example, they must, with each content created, indicate that it was generated by an artificial intelligence and provide the design of the model. The EU takes care here to fight against illegal content and to protect copyright.
The EU, ready to ban AI that would be harmful to the security in the broad sense of the citizens
European citizens will not be left out. They will indeed be able to file a complaint against AI systems, if they believe their rights have been violated. Users will also be able to receive explanations regarding decisions based on high-risk AI systems that have a material impact on rights “. Moreover, the European Parliament has chosen to prohibit the use of intrusive and discriminatory AI for remote and real-time (or a posteriori) biometric identification in public spaces (which is likely to pose a problem for the France), for predictive police systems (profiling, location, past criminal behavior), or to retrieve visual elements from social networks or video surveillance intended to create facial recognition databases.
Regarding high-risk artificial intelligence, MEPs decided to broaden the classification to include security, damage to health, fundamental rights and the environment, as we said above. MEPs also added to this list the AIs used to influence voters during election campaigns. As such, the recommendation systems used by social networks will also be classified in this category of high-risk AI.
AI presenting a so-called “unacceptable” level of risk for the safety of citizens will be purely and strictly prohibited. The EU thus targets systems that advocate manipulation, that exploit people’s vulnerabilities or that are used for social rating. In short, any AI considered harmful may be banned from the territory of the European Union.
Note that after a final vote by Parliament (which will take place between June 12 and 15), it is the Council that will have the difficult task of adopting the final and definitive form of this legislation. But the first steps have been taken successfully.
Source : European Parliament
6