The European Union's Parliament and Council negotiators, as well as the firms that produce and sell generative AI technologies, agreed new restrictions for AI software and services. The EU has become the first major international power to enact such restrictions.
The EU's Artificial Intelligence Act will prohibit the use of AI for the following:
These EU legislation will make some exceptions for police enforcement. AI systems can be used to locate particular individuals who have been victims of kidnapping, sexual exploitation, or human trafficking. They may also be used to assist prevent a terrorist attack, and law enforcement personnel can utilize them to discover suspects in crimes such as kidnapping, murder, and other crimes.
General-purpose AI systems must reveal summaries of the content they utilize for training, as well as how they comply with EU copyright regulations. They must also supply technical documentation. High-impact AI systems will be subject to stricter regulations, including measures to prevent systemic risks and reporting on their energy efficiency, among other things.
Companies found in breach of these new restrictions face fines ranging from 7.5 million to 35 million euros, depending on the severity of the offense and the size of the company.
The newly adopted regulations must be approved by the European Parliament's Internal Market and Civil Liberties committees, followed by votes by the Parliament and Council.