REVISTA DYNA NEW TECHNOLOGIES

As The Algorithm says in its presentation, it only aims to demystify artificial intelligence technologies because finally, two and a half years after its inception, an agreement has been reached: it will be the world's first AI law. It is expected to mitigate the harms that AI can cause in areas where it poses the greatest risk to fundamental rights, such as healthcare, education, border surveillance and public services, as well as prohibiting uses that pose an "unacceptable risk".
1. The Act introduces important and binding rules on transparency and ethics.
It imposes legally binding rules requiring technology companies to notify people when they interact with a chatbot or biometric categorisation or emotion recognition systems. It will also require them to tag deepfakes and AI-generated content, and to design systems so that AI-generated media can be detected. This is a step beyond the voluntary commitments made in the US. Essential AI services, such as insurance and banking, are also required to conduct an impact assessment on how the use of their systems will affect people's fundamental rights.
2. AI companies still have a lot of room for manoeuvre
The Act will require powerful underlying models and the AI systems built on them to improve documentation, respect copyrights and share more information about the data on which the model has been trained, reporting on its safety and energy efficiency. The problem is precisely the definition and measurement of power and this seems likely to change as the technology develops.
3. The EU will become the world's first AI police force
A European AI Office will be set up to monitor compliance, implementation and enforcement, the first in the world to enforce binding rules on AI, with an independent panel of experts and the possibility of high fines. User complaints can be submitted to it. They surpass US standards and have the potential to become a global benchmark.
4. It primarily defends national security
Of the four different risk groups considered, the so-called unacceptable ones are completely banned. They are those used to manipulate people's behaviour, social assessment based on beliefs or ethnicity, emotion recognition in the workplace and education, and predictive policing unless approved by the courts and only for 16 specific crimes (terrorism, human trafficking, paedophilia or drug trafficking, for example). Once in force, technology companies will have two years to implement the rules, but bans on unacceptable uses of AI will apply after six months and companies developing new models will have to comply within one year. Military use to defend national security is exempted from the law.