US Big Tech plays nice with Brussels on AI rules

BRUSSELS — Everyone expected turbulence. It became smooth sailing instead.

Brussels has secured widespread support from U.S. Big Tech and other AI leaders for its General-Purpose AI Code of Practice, a voluntary tool to ensure compliance with the EU’s landmark AI Act.

As of Friday, almost all companies in scope had fallen in line ahead of Aug. 2, a new key date in the rollout of the EU’s artificial intelligence rule book, which becomes fully applicable on the same date in 2026.

From Saturday onwards, the developers of leading AI models such as OpenAI, Google and Microsoft will have to live by a set of rules on transparency, copyright and safety. By opting for the Commission’s code of practice, they are choosing to play nice with Brussels.

It’s a win for the European Commission at a time when the bloc’s tech rule books are under pressure from the U.S.

Of all those who fall under the scope of the code, Meta is the only major holdout that refused to sign on.

Meta aside, this broad support has not always been a given, particularly during the lengthy drafting process that lasted close to a year.

Difficult assignment

The rise of generative AI, driven by the popularity of ChatGPT, prompted EU legislators to include a specific chapter in the AI Act aimed at general-purpose AI models, which can perform tasks such as generating text, videos, images and code.

The rule book provided a general sense of direction, but the specifics were left to a group of 13 academics.

They were tasked with detailing how companies would publicly disclose which data was used to train their model, their planned compliance with EU copyright law, and how they would deal with the most serious risks such as loss-of-control scenarios.

It was a difficult assignment as industry, lawmakers and civil society groups were continuously alert to ensure the guidance matched what was said in the legal text.

Midway through their efforts they also faced a change from the U.S. administration, which called upon European countries to opt for light-touch AI regulation.

That empowered companies like Meta and Google to publicly criticize early drafts of the text.

In February, Google’s top lobbyist Kent Walker slammed the set of voluntary rules as a “step in the wrong direction,” while his Meta colleague Joel Kaplan warned it went “beyond” what was laid down in law.

But months later the tone has shifted somewhat.

Walker said this week that Google would sign the code, stating that the final version “comes closer to supporting Europe’s innovation and economic goals than when it began.”

OpenAI said that signing the code reflected the company’s commitment “to providing capable, accessible and secure AI models to Europeans.” Microsoft executive Nanna-Louise Linde said the company wanted to demonstrate “long-standing compliance with EU law.”

Only Meta didn’t change its tone, with Kaplan slamming the “EU’s inconsistent, restrictive and self-defeating approach to AI” in a post on LinkedIn this week.

The Commission had warned companies that those that didn’t sign the code would face more scrutiny, as they would still have to comply with the general requirements of the AI Act.

Some companies suggested that signing the code would allow them to maintain an open dialogue with the Commission’s AI Office.

Google’s Walker said they want to work with them to ensure that “the Code is proportionate and responsive to the rapid and dynamic evolution of AI.” Microsoft’s Linde said they want to “contribute ideas for [the code’s] improvement.”

Not over yet?

One of the lingering concerns for companies is disclosing which training data was used, since that could empower rights-holders to claim remuneration from AI companies.

Walker warned against “departures from EU copyright law” and “requirements that expose trade secrets.”

xAI slammed the copyright provisions as “clearly over-reach” and refused to sign that chapter of the code of practice.

It’s an indication that next year could see new tensions between Big Tech and Brussels, with the Commission planning a review of its copyright law by mid-2026.