The European Union is one step closer to adopting their AI Act, a world first, after approving a provisional agreement to regulate the rapidly emerging technology.

The AI Act is set to regulate foundation models and generative AI, from their development to their deployment, to mitigate risks associated with the advanced tech.

It mainly focuses on large models that are trained on a plethora of data to generate outputs and perform functions, including OpenAI’s ChatGPT.

High-impact, general-purpose AI models as well as high-risk AI models will be regulated, and will have specific transparency obligations to follow. They will also be required to uphold EU copyright laws.

AI models and systems will be rated on a scale of their associated risks as far as their data harvesting and deployment.

Decision-making AI models will face some of the greatest scrutiny under the bill, and some deployments will be greatly restricted under the act, such as real-time biometric surveillance in public spaces.

High-risk applications include deployments in critical infrastructure, including health, law enforcement, education, and border patrol.

Recommended reading

• The EU Agreed on an AI Act, What Now?
• Where Will Companies Look to For AI Regulation Guidance?
• EHRC: UK’s Proposed AI Regulations “Do Not Go Far Enough”

However, the EU expects that most services will fall into low risk categories, which would include spam filters.

“I welcome the overwhelming support from the European Parliament for the EU AI Act, the world’s first comprehensive, binding framework for trustworthy AI. Europe is now a global standard-setter in trustworthy AI,” EU industry chief Thierry Breton said.

All that is left for the law to come into effect is for lawyers and translators to go over the bill, and for the European Council to give its seal of approval.