Safety researcher urges OpenAI and Anthropic to actually pace the frontier, not just talk

mealreplacer · x · 2026-11-21

Julian Hazell argues in a new blog post that OpenAI and Anthropic are all talk on "pacing the frontier." His action list: publicly engage with AIFP's August pacing plan, reallocate staff urgency like they did for security, stop redacting so much of METR's report (the full version could persuade skeptics), give OpenAI's embedded evaluators comparable access, and far more than hiring a single pacing TPM. Given both companies' own short timelines to dangerous AI, he says they should hurry up and act.

Related event: Researchers Urge OpenAI and Anthropic to Act on Slowing the Frontier(2 posts)→

Original post →

More from Companies & People

Companies & People channel →