What's this about?
More than 1,100 employees from OpenAI, Anthropic, and Google publicly signed a statement asking the U.S. government to help build the technical and governance tools needed to "deliberately pace" how fast frontier AI develops — especially the part where AI starts accelerating its own development. Unusually, both OpenAI and Anthropic posted official corporate support the same day.
What exactly does "pacing the frontier" mean?
"The frontier" is shorthand for the most advanced AI models being built right now. "Pacing" it means deliberately slowing down — or at least keeping the ability to slow down — the rate of progress, so society, regulators, and safety research can keep up. The specific fear named in the statement is "recursive self-improvement": AI systems getting good enough to speed up AI research itself, which could make progress lurch forward faster than anyone can react to. It's less "stop building AI" and more "build the brakes before we need them."
What's the community saying?
- Sincere support: people who work at labs saying they take the pace seriously and that HN's reflexive "they're just protecting their moat" dismissals of Dario/Demis/Sam are lazy and non-constructive.
- The "it's naive or bad-faith" camp: has technology ever voluntarily stopped advancing worldwide just because people wanted it to? And "China won't slow down for you," so unilateral pacing just cedes the lead.
- The "then quit" jab: if you truly believe the work is dangerous, the consistent move is to stop doing it — critics note that's conspicuously not what signatories are doing.
- Vagueness: the actual ask ("support an international effort to develop technical and governance tools") is so unspecific that readers couldn't tell what concrete action is being requested.