LEAK: Given the massive public attention to the #AIAct, I've taken the rather unprecedented decision to publish the final text. The agreed text is on the right-hand side for those unfamiliar with four-column documents. Enjoy the reading! Some context: the document was shared with EU countries yesterday afternoon, ahead of a discussion within the Telecom Working Party, a technical body of the EU Council, on Wednesday and formal adoption at the ambassador level (i.e. COREPER) on 2 February. Therefore, the timeline is rather tight, and national delegates will not have enough time to analyse the entire text but will have to focus on the key articles. France is still testing the waters with other countries on the possibility of forming a blocking minority, at least to delay the COREPER vote, with the view of getting some concessions into the text. At COREPER, European governments will be asked to give a yes or no answer, as the Belgian presidency is not accepting any comment or reservation at this advanced stage. So far, Paris has been unable to build a blocking minority, but the picture will become clearer when the member states provide their feedback at the technical level. But if France does not get any concessions at this stage, it will keep up the pressure to influence the AI law's implementation, notably in terms of secondary legislation, as the file remains a top national priority. https://lnkd.in/etdMhw4c
Thanks for sharing. It's difficult for an American like me to understand why this needs to be "leaked." The European Commission voted and agreed upon the EU AI Act in principle. But the details still aren't fully worked out and the text isn't (officially) public. How does this work? In the US Congress, every bill voted on is publicly available.
Disliking the insertion of 'human-centric' as industry frequently interprets exactly the opposite meaning of this phrase than what regulators intend
OMG, it's 892 pages! Who is going to read it? I will definitely rely on a GenAI app to summarize it, taking into account that it will not be 100% reliable. I thought it had already been accepted by the EU :) Perhaps the third time's the charm. The EU law enforcement process is so lengthy that by the time a regulation becomes valid, there's often no longer a need for it😊
Blanks on the right hand column mean no agreement yet?
So Art. 28b as proposed by the EP is out, as part of the final text? But materially partially still there? Philipp Hacker Dr. David Bomhard Kai Zenner
Why did I feel that i am a bit shivering...#GDPRmoment
Many thanks to Luca Bertuzzi for your invaluable contribution. It's clear to see that it's greatly appreciated by the public! For those feeling overwhelmed by the AIA, we've developed a custom GPT to simplify and clarify its complexities. Access it here: https://www.wearelegit.ai/aia-gpt (ChatGPT Plus required). Our GPT encompasses all relevant legal documentation, in addition to other knowledge base materials such as timelines, FAQs, and a summary of the institutional context of the EU AI Act by Hadrien Pouget, an AI policy expert at the Carnegie Endowment for International Peace. Thanks to Luca Bertuzzi, we have now also updated our knowledge base in line with the four-column document. We also acknowledge that the AI Act GPT may give false answers, so we value any feedback to fine-tune settings and avoid critical errors. Thank you! 🧩
France already won as it force the discussions on regulations on the foundation models from the “how” to the “it”. Inline with the limited mandates of the EU, regulation of the most afvanced AI for safety and unaccountable concentraiton will be left to the US government and national security agencies, with some consultation of our agencies. Same as it alsways been for dangerous and potent techs from nuclear, bio weapons and encryption. Hopefully we can do better this time, but that requires a global constituent assembly that can create IGOs powerful enouhh to truly manage, tame and share this techs for the interest of all. We are giving it a shot with our Harnessing AI Risk Initiative: https://www.trustlesscomputing.org/harnessing-ai-risk-proposal
Technology journalist
4moA clean version is now available https://www.linkedin.com/posts/dr-laura-caroli-0a96a8a_ai-act-consolidated-version-activity-7155181240751374336-B3Ym?utm_source=share&utm_medium=member_desktop