According to the new rules, developers must detect how their models work and what data used for training, the French Le Monde conveyed.
For more advanced models, which carry larger risks, security measures must be documented.
These rules allow for better copyright protection, as developers must report data sources used, including possible automation of data collection from the Internet.
Also, measures must be applied to protect intellectual property.
The European Commission 27 Member States must notify which national authorities will be responsible for ensuring the proper implementation of regulations.
This law causes tensions between the EU and the United States, because US companies, including Google and target, are not enthusiastic to regulations that could slow down the development of AI technology in Europe.
Although the EU issued a good practice code, many large American companies oppose the laws that undertake them to greater supervision and responsibility.
Developers who violate the law may be punished with up to seven percent of global income.
The law also includes provisions for high-risk AI systems in sensitive sectors, such as education, energy infrastructure and border control.
(Vijesti.ba)
