Summary:
Les principales entreprises d’IA, comme OpenAI, Microsoft et Google, disposent désormais d’un guide pour se conformer à la Loi sur l’IA de l’UE. Cette conformité, qui inclut la transparence des données d’entraînement, peut influencer les régulations mondiales. Bien que le code de pratique soit volontaire et vise à alléger la charge administrative des entreprises signataires, certains estiment qu’il impose un fardeau disproportionné sur les fournisseurs d’IA, ce qui pourrait affecter leur compétitivité.
Original Link:
Original Article:
The world’s leading AI companies now have a playbook for how to comply with Europe’s AI Act, thanks to guidance released by the bloc last week.
**Why it matters:** How providers of general-purpose AI models like OpenAI, Microsoft and Google adhere to the EU’s AI laws will be revealing for the rest of the world as governments attempt to understand and regulate AI.
* The compliance, and what companies are required to demonstrate, will highlight some of the inner workings of AI companies that have mostly been secret — like the content used to train AI algorithms.
* Companies that voluntarily sign onto this code will “reduce their administrative burden,” according to the European Commission.
* Whether a company decides to sign may also indicate how much lobbying influence it was able to have over EU regulators’ decisions.
**Driving the news:** The [code of practice](https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai) is a voluntary guide meant to help companies show they’re complying with the AI Act, broken into three chapters: transparency, copyright, and safety and security.
* Compliance involves releasing summaries of AI training data, mitigating use of copyrighted content and using internal risk frameworks.
* The EU’s rules for general-purpose AI systems take effect on Aug. 2, but companies have until August 2026 until penalties or fees kick in.
**What they’re saying:** Companies and coalitions have mixed reviews of the code of conduct so far. OpenAI said it intends to sign it and that it has plans to double down on its European business.
* “Too often in Europe, the limelight has been taken by regulation,” an OpenAI blog post [reads](https://openai.com/global-affairs/eu-code-of-practice/). “Now it’s time to flip the script and use this moment to empower Europe’s innovators to innovate and builders to build for Europe’s future.”
* The company still hopes it gets simpler, though, for the sake of startups and smaller firms: “We have [advocated](https://openai.com/global-affairs/openais-eu-economic-blueprint/) for greater simplification and harmonization to support these next generation companies and will continue to back their concerns, as they are key to AI of, by and for Europe.”
**The other side:** CCIA’s European office said the code of practice “imposes a disproportionate burden on AI providers.”
* “Without meaningful improvements, signatories remain at a disadvantage compared to non-signatories, thereby undermining the Commission’s competitiveness and simplification agenda.”
* Business Software Alliance director of EMEA policy Hadrien Valembois said in a statement the code had made “welcome improvements,” but pushed for companies to have more time to comply.
* Microsoft declined to provide a comment. Meta, Google and Amazon didn’t respond to requests for comment.
**The bottom line:** The European Commission first published a proposal for regulating AI in 2021, far before the explosion of generative AI.
* Amid sensitive trade discussions with the Trump administration, the threat of tariffs, and the U.S. pivoting from a safety to beating-China-at-all-costs attitude toward AI, EU regulators have reason to listen to U.S. companies that want to do business in Europe.