The European Union has become the first supranational body to establish a legal framework for the development of Artificial Intelligence. Approved on December 8, it is the first time that the use of AI will be allowed or prohibited based on the risk to humanity. The measure also aims to boost the development of AI in Europe against the push of giants such as the United States or China.
The legislation still has to be ratified by the European Parliament and the Council of the European Union, but its approval represents an important step forward and demonstrates that an agreement could be reached despite the obstacles that seemed insurmountable, with compromised issues. such as the use they could give to the biometric identification cameras security forces to prevent terrorism or public disorder.
Finally, biometric categorization systems that use sensitive information, such as religious beliefs, sexual orientation or race, may not be used. Nor are cameras that collect facial images indiscriminately on the Internet or in public, except in specific cases with judicial authorization and with the express safeguard of human rights. Also when they are applied only to one person – for example, if they are suspected of having committed a serious crime – or to search for victims of kidnapping or sexual exploitation.
The regulation of Artificial Intelligence models – such as ChatGPT, Google Bard or Google Gemini – also focused the negotiations. They are seen as an “opportunity” but at the same time it is “fraught with risks.” These models will have to meet strict criteria and carry out evaluations of their models, pointing out systemic risks and working to minimize them. In addition, companies will be obliged to report to the European Union about any problems they suffer and ensure the cybersecurity of citizens.
As a general rule, the rule details that all those that pose an unacceptable risk to human beings and democracy will be prohibited. In this sense, applications that recognize emotions in the workplace or in educational centers are also prohibited, as well as social scoring that allows discrimination against the individual in access to facilities, tasks or public aid. Also Artificial Intelligence systems that can be used to exploit vulnerable people.
The rule is expected to come into effect by 2026, once it has been ratified by the Parliament and the Council of the EU. Failure to comply with the legislation will lead the European Union to impose very severe fines on AI companies that break the rule. These sanctions can range from 7.5 million euros to 35 million euros; or from 1.5% of its worldwide turnover to 7% of total turnover.
