EU negotiators have agreed a deal on the world’s first comprehensive rules for artificial intelligence.
The agreement paves the way for legal oversight of technology used in popular generative AI services such as ChatGPT.
Negotiators from the European Parliament and the bloc’s 27 member states overcame major differences over generative artificial intelligence and police use of facial recognition to sign a tentative political agreement on the artificial intelligence law.
“Share!” tweeted EU Commissioner Thierry Breton.
The European Parliament and member states “have finally reached a political agreement on the Artificial Intelligence Act!” tweeted the parliamentary committee that helped lead the body’s negotiating efforts.
Officials have provided few details about what will make it into the final law, which won’t take effect until 2025 at the earliest.
The EU took an early lead in the global race to draw up AI safeguards when it unveiled the first draft of its 2021 rulebook.
However, the recent boom in generative artificial intelligence left European officials scrambling to update a proposal poised to serve as a blueprint for the world.
Generative AI systems such as OpenAI’s ChatGPT have become increasingly ubiquitous in recent months – impressing users with their ability to create text, photos and songs, but also causing concerns about jobs, privacy and copyright protection.
Now the US, UK, China and global groups such as the G7 have jumped in with their own proposals to regulate AI, although they are still catching up with Europe.
When the final version of EU‘s AI law has been drawn up, the text must be approved by the bloc’s 705 politicians before they break up for EU-wide elections next year. That vote is expected to be a formality.
The AI Act was originally designed to mitigate the dangers of specific AI functions based on their level of risk, from low to unacceptable.
But policymakers pushed to extend it to base models, the advanced systems that underpin general-purpose AI services such as ChatGPT and Google’s Bard chatbot.
What became the most difficult issue was AI-powered facial recognition surveillance systems, and negotiators found a compromise after intensive negotiations.
European politicians wanted a complete ban on public use of facial scanning and other “biometric remote identification” systems due to privacy concerns, while member state governments wanted exemptions so law enforcement could use them to tackle serious crimes such as child sexual exploitation or terrorist attacks.