Technology AI, Science and New Things
The EU have reached agreement on the details of their landmark AI act legislation, after 37 hours of gruelling negotiations. The bill will be the world’s first comprehensive AI law, introducing regulations on everything from the use of facial recognition systems to new Generative AI systems, with a focus on mitigating risk to humans.
Negotiations had reached an impasse between the EU parliament and member states over how to regulate powerful AI “foundation” models like OpenAI’s GPT-4 – France and Germany had pushed for soft self-regulation to project their own AI startups.
On Friday negotiators finally reached a compromise, introducing transparency and harm-reduction measures on Generative AI systems, but exempting open-source models (models released publicly for others to use or modify freely) – a win for France’s Mistral AI, recently valued at $2 billion, and Germany’s Aleph Alpha. While certain corners of Silicon Valley have derisively characterised the bill as Europe regulating what they cannot build themselves, US congress will likely introduce their own AI regulations in the near future, and will look to the precedent the EU have now set.
