Most of the world’s biggest AI players have bitten the bullet and signed up to the EU AI code of practice despite their misgivings, a list of signatories published over the weekend has revealed.

And end user firms have been warned that they too could get caught in the EU’s net after the trading bloc’s AI Act came into force.

Obligations for the biggest AI models kicked in on August 2, as the General-Purpose AI Code of Practice officially came into effect.  Brussels issued the code of practice for organisations providing the biggest models in early July and followed up by publishing guidance for providers a week later.

At the time, EU sources said that it was down to companies to decide whether to sign up to the code of practice. But if they chose not to, they would have to demonstrate other ways that they were meeting their obligations under the AI Act.

Over 25 firms have now signed up, including Microsoft, Google, Open AI, Mistral AI and ServiceNow, with more likely to come.

But Meta has held back from the entire code, previously saying it was too restrictive and set Europe “down the wrong path.” xAI has signed up to the chapter on safety and security but not the rest of the code, accusing it of “overreach”.

And the list did not feature any Chinese-based players at all.

Wayne Cleghorn, a partner at Excello Law, who was part of the working group that shaped the Code of Practice, said today this was not the end of the story and that every organisation needed to consider how the Brussels legislation would affect their Ai strategy.

“The final text represents a consensus built under real time pressure. It’s a best effort starting point, and while it will need to evolve, it establishes foundational norms for AI governance.”

And, he added, the enforcement regime will now start taking shape, as the individual member countries put oversight, compliance and penalty rules in place.

Cleghorn said other businesses should now assess whether their AI systems fell under the law, including British firms – despite the fact that the UK is not a member of the EU. “UK companies, including those building or deploying AI systems that enter the EU market, could be directly impacted.

“It’s crucial to assess whether your AI systems fall into categories of ‘unacceptable’, ‘high’, ‘limited’, or ‘minimal risk’ as defined under the Act, and then apply the relevant compliance obligations.”

Businesses needed to get “a firm grasp” of the systems they develop, deploy “or use” and how the act applies to them, Cleghorn said. “In most cases, this requires assembling multidisciplinary internal teams and seeking targeted legal and regulatory advice. Staying ahead of the evolving rules is no longer optional.”

The EU’s rules do extend to downstream companies who develop their own cut-down models or systems based on models developed by big tech.

Back in July, commission insiders confirmed that the threshold for general purpose models was 10^23 FLOP.

Modified models developed using one third of this threshold of more would also fall under the act, they said.

But, they added, the commission expected that threshold “will rarely, if at all, be met, so we think it will be a very, very few cases.”

The Commission said in a statement on Friday, “Providers who sign and adhere to the Code will benefit from reduced burden and increased legal certainty.” Models on the market before August 2 have until August 2 2027 to ensure compliance.

Google’s president of global affairs, Kent Walker confirmed last week that it would sign the code, but expressed concerns that the AI Act and Code risked slowing Europe’s development and deployment of AI.

“In particular,” wrote Walker, “departures from EU copyright law, steps that slow approvals, or requirements that expose trade secrets could chill European model development and deployment, harming Europe’s competitiveness.”

Microsoft’s VP for European government affairs, Nanna-Louize Linde, wrote on LinkedIn that it was signing up to the code of practice, to further build trust in its AI models.

But, she added,: “The AI Act is a complex regulation that would benefit from simplification.”

Personalized Feed
Personalized Feed