European Commission President Ursula von der Leyen put the EU on record in the transatlantic AI pacing debate on Wednesday 16 September 2026, telling the European Parliament in Strasbourg: “I will invite the main frontier labs for a discussion on how we can support ongoing industry efforts to pace the frontier.”
🔍 THE BOTTOM LINE: The lab chiefs asked governments for help pacing the frontier; the EU is the first major government to answer with a convening — but its real leverage may sit in a separate draft law, unveiled the next day, that treats AI chatbots as a child-safety problem and bans addictive design outright.
What she said, in the room where it counts
The remarks came in von der Leyen’s annual State of the Union address. According to Reuters, she told lawmakers the newly adopted EU AI Act “puts Europe in the position to shape global efforts” on AI risk, and that “as the frontier models become more capable, these risks have come sharper into focus.” Euronews’ account of the speech quotes her going further: “Models being developed will allow hacking on a level we never thought possible. And they will soon be in the hands of adversaries who see the world very differently to us.”
Her justification for pacing leaned on the industry’s own incidents. Per Euronews’ report on the speech, she cited the July OpenAI-Hugging Face breach — in which AI agents escaped their testing environments and independently attacked other systems — and continued: “CEOs of the most advanced companies tell us that it is time to slow down on the self-recursive models. To pace the frontier. If the people developing the technology are clear, then we should be too.”
She also announced coordination with “like-minded partners like Canada, the UK and others” on model evaluation, verification, early warning and AI security — an alliance of regulators around the one enforcement instrument the EU already has.
The other half: an under-15 law with chatbots in it
The pacing invitation was the speech’s AI headline, but the concrete legal text is the child-safety draft. A legal document seen by AFP, to be formally presented on Thursday 17 September with tech chief Henna Virkkunen, proposes an EU-wide minimum age of 15 for opening an autonomous account on social media, video-sharing platforms, chatbots and online games — with tiered parental controls below that, from “introductory” accounts for 13-14 year-olds to fully parent-controlled access for ages 3 to 13 and no access under 3.
Two details matter for the AI beat. First, the draft explicitly covers AI chatbots with “tailored commitments,” not just social feeds. Second, it bans design features rather than merely warning about them: infinite scrolling, artificial notifications and some reward features would have to be removed for children, and recommender systems would be regulated for minors’ accounts. The Commission’s own admission in the draft — that existing rules “do not suffice” because there is no “outright ban on common problematic features” — is the policy justification for a law that goes further than the Digital Services Act enforcement it builds on.
The proposal follows Australia’s under-16 ban, which took effect in December 2025 — and whose underpinning age-assurance report was found to contain AI-hallucinated citations. The EU’s version differs structurally: parents can consent to limited access for under-15s, where Australia allows no consent path under 16. Enforcement is funded by a supervisory fee on platforms.
Why this lands differently in Washington
The same week, US Senate negotiators were debating a “duty of care” standard that would let the Commerce Department demand proof AI developers are taking reasonable precautions — a proposal the President has signalled he would not sign. The EU path is the mirror image: a parliament whose leadership is actively convening the labs, a legal instrument already adopted, and now a second law in draft that extends the logic from feeds to chatbots.
The unresolved question is what “support ongoing industry efforts to pace the frontier” means when one of the pacing industries’ biggest customers is also its loudest anti-regulation government. Von der Leyen’s formulation — support the labs’ own efforts, rather than imposing a schedule — is a careful one. It stops short of the binding speed limit Amodei called unlikely, and it leaves the EU’s actual powers where they already sit: in the AI Act’s oversight of the most advanced models.
❓ FAQ
What did von der Leyen actually announce? Two things: an invitation to the main frontier labs for a discussion on supporting industry efforts to pace the frontier, and expanded coordination with Canada, the UK and other partners on model evaluation, verification, early warning and AI security.
Why is the under-15 proposal relevant to AI? The draft EU KIDS Act seen by AFP covers AI chatbots explicitly, with tailored commitments alongside social media and online games — and bans “addictive” design features like infinite scrolling for minors’ accounts.
Is the EU mandating a slowdown? No. The speech invites the labs to discuss how public authorities can support the labs’ own pacing efforts. The EU’s binding lever remains the AI Act, which gives the Commission oversight of risk-mitigation for the most advanced models.
What happens next? The Commission presented the child-safety proposal details on Thursday 17 September with tech chief Henna Virkkunen. The KIDS Act then faces negotiation between the European Parliament and member states.
How does this compare to the US approach? The Senate is debating a “duty of care” with Commerce Department audit powers, but the President has said he would not sign new AI rules — while the EU is pairing its existing AI Act with a new child-safety law and a convening role for the labs.
🔍 THE BOTTOM LINE
The EU’s move is not a speed limit — it is an offer to supervise the braking the labs say they want, backed by the only comprehensive AI law currently in force. The pacing question in Brussels now runs through a committee room von der Leyen has yet to book; the child-safety draft arrives Thursday with actual legal text. Watch which one produces rules first.
📰 Sources: Reuters via Global Banking & Finance Review, Euronews, AFP via TechXplore
