Listen: How does Trump’s AI pact compare with the EU?

Production: By Europod, in co-production with Sphera Network. EUobserver is proud to have an editorial partnership with Europod to co-publish the podcast series “Briefed” hosted by Léa Marchal. The podcast is available on all major platforms. Find the full transcript below: Donald Trump is betting on voluntary commitments from AI companies. On Tuesday (29 September), he unveiled a self-regulatory agreement drawn up with some of the biggest names in the AI industry. What does it mean, and how does it differ from the EU’s Code of Practice for general-purpose AI? On Tuesday, September 29, Trump brought together the CEOs of some of the biggest AI companies. Together, they signed what could be described as a self-regulation pact: a short list of non-binding commitments called the White House Accord on Super Intelligence. That’s the other novelty: Trump no longer wants to talk about artificial intelligence, but about “super intelligence”. The term AI apparently has negative connotations. So, what exactly have these companies committed to? The CEOs of OpenAI, Anthropic, Google, Meta, SpaceX and Nvidia have pledged to put safeguards in place, quickly detect and fix problems, and have their systems checked by independent auditors. The principle of voluntary commitments may sound familiar to Europeans: They echo the EU’s Code of Practice for general-purpose AI — which, like the White House agreement, is not legally binding. So what’s the difference? First of all, OpenAI, Anthropic and Google have also signed the EU code, while Meta, SpaceX and Nvidia have not. And, more importantly, the EU does have binding rules in place: the AI Act, which is already in force. As part of this regulation, the European Commission developed the Code of Practice to help companies comply with the rules. You could think of it as a checklist for AI companies, to help them make sure they are complying with the law. Another difference is the sheer volume of recommendations. The document published by Trump is just one page long. The European Code of Practice runs to more than 60 pages and was prepared with the help of dozens of independent experts. It is also more comprehensive, covering not only safety, but also transparency and copyright. But this raises another question: if both documents are voluntary, how effective can they actually be? When the European Code of Practice was published, some civil society organisations criticised it, arguing that voluntary commitments are not enough and that concrete obligations are needed. There are binding obligations under the AI Act. But they are not as precise as what is in the code of conduct. Similar criticisms have emerged since the White House agreement was published on Tuesday. And it is worse in the US case, since companies are absolutely not legally bound by the commitments they have made. The United States has generally been less inclined towards binding regulation of the sector. Trump has argued that rules can hinder the development of companies, and has promoted a national framework that is as minimally restrictive as possible. The White House document does leave the door open to legislation in the future. It says that, over time, it may make sense to codify these steps into lows or regulations. But for now, they remain voluntary commitments. If you are interested in this topic, there will be a live event on Tuesday next week about AI and its safeguards, organised by VoxEurope. It will be online and you can find all information on Vox Europe website.