Slowing down...and then what

The argument over whether to “pace the frontier” says more about how we see power and responsibility than about kilowatts and GPUs. Dario Amodei’s call for the major labs to voluntarily pace their most advanced work — coupled with independent third‑party safety evaluators, common standards and some international buy‑in — is a pragmatic attempt to replace the false binary of full throttle or full stop. It’s telling that Sam Altman and even Elon Musk have signalled sympathy for some version of this; they can read the political weather as well as the technical tea leaves. From a European vantage point — where the EU AI Act is moving from text to texture — the language matters: “pace” invites stewardship, whereas “slow down” can sound like abdication.

Still, there’s a hard truth beneath the rhetoric: open‑source communities will not pause, Chinese labs will not wait, and the knowledge required to push capabilities forward is already diffuse. Trying to slow global research with only voluntary pledges is like asking the weather to be reasonable. The better question is where pacing changes the risk surface without pretending to control the sky. That’s where independent evaluation, model reporting and coordinated safety benchmarks can help, not as a moral victory but as risk plumbing — closer to aviation incident reporting than to grand bans. Europe has muscle memory here: standards bodies, conformity assessments, liability regimes — tools that can lend credibility if they are genuinely independent, funded, and capable of keeping up.

For business leaders, the practical posture is neither accelerationist bravado nor regulatory fatalism; it’s disciplined readiness. Treat frontier pacing as an opportunity to insist on evidence before exposure: demand real evaluations, insist on transparency around failure modes, and wire your operations to absorb rapid model upgrades without betting the firm each quarter. At the same time, keep one eye on the outside lane: open‑source models and non‑Western ecosystems will keep compounding, and your competitiveness depends on learning to integrate them safely rather than wishing them away. In other words, pacing is tempo, not timeout — a chance to breathe, assess, and then move with deliberate speed toward a future we help shape, rather than one that simply happens to us.