Artificial Intelligence models must announce their educational data in Europe

EU. By introducing New artificial intelligence instructionsAsked companies active in this area to transparently Educational data Disclose their models and from violations Right to publish Avoid.

According to reports, the move is part of the union’s extensive efforts to implement the first comprehensive legal framework of artificial intelligence regulation in the world, introduced last year. Although these laws have not yet been fully enforced, recent guidelines to facilitate their implementation by technology giants such as Openai, Microsoft And Google Has been provided.

New Artificial Intelligence Instructions in the EU

New artificial intelligence instructions to help Long Language Models (LLM) Provided to adapt to new rules. This law from August 2 (August 11) It is implemented, but the new models have one year and current models have two years to adapt. The instructions are set in three key parts of transparency, publication and safety.

In the field TransparencyCompanies are required to provide accurate information on how to train models, energy consumption, processing power and educational data resources, which will probably face corporate resistance.

Section Right to publish He emphasizes that language models must respect the rules of publishing, including Paywalls and the restrictions on the crawling of websites. This has caused disputes; European publishers, for example, have complained about the feature of Google’s artificial intelligence. Denmark has also suggested a law that gives individuals the right to publish their image to prevent the production of unauthorized diplomas.

Companies in the field of artificial intelligence have accepted that their technology may be incompatible with the laws of the current publishing, and some executives have warned that the requirement to obtain permission from all owners of publishing may stop artificial intelligence activity suddenly.

In the field SafetyArtificial intelligence guidelines refer to potential threats, including violations of personal security, extensive oversight, weapons development, fraud, and inaccurate information. According to Bloomberg, violations of the European Artificial Intelligence Act could penalize up to 7 % Get out of the annual income of companies that can have an important deterrent effect.