The AI Act needs to be voted on by the 27 countries in the EU Parliament|@DefendDemocracy|X
The European Union (EU) became the first governing body in the world to reach a provisional deal on comprehensive laws to regulate artificial intelligence on Friday.
Though the deal needs to be voted on by the 27 countries in the EU Parliament and will probably go into effect only in 2025, it still positions Europe as a leader in AI tech regulation.
Most of the legislation focuses on what AI tech companies, including ChatGPT’s OpenAI and Google’s DeepMind AI, need to do when operating in the EU. Here are some,
-
Data disclosure and rigorous testing on AI technology used in high-risk applications like self-driving cars and medical equipment.
-
Ban on scraping faces from the internet or security footage to create facial recognition databases.
-
Labeling AI-generated images, similar to what the October AI executive order from President Joe Biden proposed.
-
Restrictions on foundational models that are classified as having “systemic risk.”
-
Relaxations on open source models, which in turn could allow developers to freely alter code for their use. It may even benefit Meta, which recently launched its open-source model LlaMA.
-
Financial penalties of up to 7% of the global revenue for the companies that violate the AI Act.
Critics of the AI Act argue that the law would hinder AI developments in the EU, giving an advantage to America and the UK in research and development.