The United States warned the European Union that its proposed AI regulations would benefit companies with the resources to shoulder compliance costs while hurting smaller ones, according to previously unpublished documents.The U.S. analysis is mainly based on the European Parliament’s version of the Artificial Intelligence Act, which includes regulations related to generative AI. Documents show that some of the rules of the European Parliament's version of the artificial intelligence law are based on "vague or lack of definition" provisions.
The analysis is Washington’s most detailed position on EU AI legislation.One issue that worries the United States is that the European Parliament is concerned about the way AI models are developed, while the United States is more inclined to focus on the risks involved in the actual use of these models.
The analysis warned that EU AI regulations could dampen AI's expected boost to productivity and could lead to a shift of jobs and investment to other markets.
The analysis said the new rules could also hinder "EU investment in AI R&D and commercialization, thereby limiting the competitiveness of European companies" because large language model training relies on intensive resources.
In 2021, the European Commission proposed the world's first draft law to regulate artificial intelligence. Since then, the European Parliament and the European Council have conducted multiple rounds of revisions and discussions on the draft. After the emergence of generative artificial intelligence applications such as ChatGPT, EU legislators urgently discussed issues not covered in the original draft.
According to the draft, the EU should divide artificial intelligence systems (such as ChatGPT, etc.) into different risk groups. The more dangerous the application, the stricter the rules that apply.At the same time, the EU should also enact a comprehensive ban on artificial intelligence, such as a ban on interactive conversation tools that encourage dangerous behavior and predictive policing tools used to predict who will commit crimes.
It is worth noting that in the latest draft displayed on the official website of the European Parliament, real-time remote biometric recognition technology is adjusted from the "high risk" level to the "prohibited" level, which means that in the future, companies will not be able to use AI technology to perform face recognition in public places in EU countries.