On 16 September, Ursula von der Leyen stood in front of the European Parliament and said something that would once have sounded like an admission of defeat: AI development needs to slow down.

In her State of the Union address, the European Commission President said self-improving AI systems should pace themselves rather than race ahead. She announced she will invite the leading frontier labs, the same companies building the models everyone is racing to adopt, to talks on how the EU can support what she called "pacing the frontier." She warned that models currently in development could enable a level of hacking that Europe hasn't previously had to plan for. And she committed the EU to joint work with the UK and Canada on evaluating and verifying advanced AI models, early warning systems, and AI security more broadly.

That last point is the one worth sitting with if you run a UK business. This isn't Brussels talking to itself. It's the EU's chief executive naming the UK as a partner in the work of checking whether AI systems do what they claim to do. For any UK agency with EU clients, EU staff, or EU-based operations, that's a direct line back to your own business, regardless of where you're headquartered.

It also lands at an interesting moment. For the best part of a week, the pacing conversation has been an industry one: labs like Anthropic, OpenAI and Google talking, among themselves, about whether the race needs a shared brake. Von der Leyen's speech is the first time a political leader with real regulatory standing has stepped into that conversation from the other side of the table. Reported alongside her remarks: naming Dario Amodei, Elon Musk and Sam Altman among those broadly aligned with pacing, and Mark Zuckerberg and Nvidia's Jensen Huang among those pushing back.

Here's the useful, calm reading of this for an agency owner, rather than the alarmist one. The headline isn't "the EU wants to slow AI down." It's that the people now writing the rulebook are saying, in public, at the highest level, something close to what gets said quietly in every Diagnostic Nathan runs: pace matters more than speed, and oversight only means something once someone with real standing is doing it. A policy of "move fast" and a policy of "verify as you go" aren't opposites. Most agencies doing well with AI right now are already running both.

That's the shape of the work Go Wisely did with The Turner Agency, a global events and film agency of around 36 people. Rather than a single crash-course rollout, TTA went through a full AI readiness diagnostic across six dimensions, including Governance, followed by a bespoke six-session AI Accelerator training programme delivered over several weeks. The pacing was deliberate: each session built on what the last one had settled, rather than trying to cover everything in one sitting. Across 131 survey responses collected over the programme, satisfaction came in at 4.6 out of 5, with not one dissatisfied response.

None of this means every agency needs a six-session programme, or that the EU's move changes anything for a business with no European customers or staff. It does mean the argument for taking AI adoption at a considered pace, rather than either freezing or sprinting, just got a lot harder to dismiss as overcaution. If a head of state is telling the companies building the models to slow down and show their working, an agency doing the same with its own team isn't behind. It's ahead of where regulation is about to land.

The practical next step, if any of this rings true, isn't to panic-audit your AI use before the EU catches up with you. It's the same first step Go Wisely recommends before any of it: a straightforward conversation about where AI genuinely fits in your business, and where it doesn't, before deciding what pace is right for you.

Go wisely.

Source: European Commission State of the Union address, 16 September 2026, reported by Help Net Security and Reuters.