Summary:
L’UE a approuvé aujourd’hui le Code de Pratique pour l’IA Générale, marquant une avancée majeure dans la gouvernance de l’IA. Ce code, élaboré après de nombreuses négociations, sert d’outil volontaire pour que les fournisseurs de modèles d’IA Générale démontrent leur conformité avec la Loi sur l’IA. Composé de chapitres sur la transparence, le droit d’auteur et la sécurité, il accorde un an de délai aux signataires avant l’entrée en vigueur des obligations. Actuellement, 26 signataires, dont Google et Microsoft, ont rejoint cette initiative.
Original Link:
Original Article:
BREAKING: EU approves the General-Purpose AI Code of Practice TODAY
In yet another landmark moment for AI governance, the EU has formally approved the GPAI Code of Practice.
After months of intense negotiation involving hundreds of experts, the final version of the GPAI Code of Practice was published by the European Commission a few weeks ago.
However, it had not yet been formally approved as an official Code of Practice that the EU AI Office recognised. That changed today.
In a brief statement, it was confirmed that the European Commission and the European AI Board have approved that the GPAI Code of Practice as an “adequate voluntary tool for providers of GPAI models to demonstrate compliance with the AI Act”.
The Code of Practice consists of three separate chapters, on transparency, copyright, and safety and security. It is complemented by European Commission guidelines on general-purpose AI, as well as an official template for publishing GPAI Model Training Data Summaries.
In classic last minute fashion, this approval comes just one day before the obligations for providers of GPAI models become applicable.
From tomorrow (2 August 2025), GPAI models made available in the EU must comply with the AI Act’s extensive provisions.
However, signatories of the GPAI Code of Practice are being granted a de facto 1-year grace period by the AI Office.
Also, providers of GPAI models placed on the market before 2 August 2025 (i.e., every GPAI model available today) do not have to comply with the obligations for those models until 2 August 2027.
The Code of Practice already has 26 confirmed signatories, including Google, Microsoft, OpenAI, and Anthropic.
If there is one thing you should not underestimate the EU on, it is their ability for successful last minute deal-making on matters of intense political and regulatory complexity.
As emphasised by the European Commission a few weeks back, there will be no meaningful delay to the EU AI Act.
Buckle up!