The Law of AI of the European Union He entered into force On August 1, 2024, and yesterday the European Commission published a statement in which the uses and prohibited practices of AI.
These “prohibited practices” are detailed in the text of the law of AI. Specifically, In its article 5. The objective is to try to cover a series of cases in which AI can try to interact with people, both through consumer applications and through physical environments.
European regulation defined four risk levels For AI systems ranging from the minimum risk (spam filters) to unacceptable risk, considered systems as a threat to safety. Among the unacceptable activities are the following:
- Ia for social scoreallowing to create risk profiles based on people’s behavior.
- Manipulators capable of affecting people’s decisions in a subliminal or deception.
- Which exploits vulnerabilities such as age, disability or socioeconomic level of users.
- The one who tries to predict whether people will commit crimes based on their appearance
- Ia who uses biometry to infer the characteristics of a person, such as their sexual orientation
- Which collects biometric data in real time in public sites for police security purposes
- Which records and tries to infer users’ emotions both in the workplace and in academic institutions
- Which creates or expands facial recognition databases collecting online images or through security cameras
Companies that violate this regulation will face fines of up to 35 million euros or up to 7% of its annual income The previous fiscal year, the figure that is greater. And those fines affect any company, regardless of where its headquarters are installed.
Last September 100 companies signed the IA Pact of the EU, a voluntary initiative to promote the application of the principles of the law of AI. Among the signatories were giants such as Amazon, Google and Openai, and all promised to identify potentially qualified AI systems as high risk.
There is nevertheless Some exceptions to those prohibited uses of AI. Thus, for example, the AI law allows security agencies to use certain biometric data collection systems – like our face – if these systems help make “directed searches” for for example victims of kidnappings or avoid imminent threats to People. That would validate uses like the one who Recently we met in Madrid.
The countries of the European Union have until August 2 to designate which authorities will monitor this market and guarantee compliance with the rules of the AI Law. The question is the impact that this will have on the availability of AI functions in the EU.
Since there is talk of high -risk systems, that impact should be minimal, but The rest of the regulatory framework -with The DMA as the protagonist— Yes propose an important obstacle so that in the EU we enjoy the latest advances in this field.
In Xataka | The EU’s artificial intelligence law was born obsolete. So you are already preparing the following