Chris Manning Proposes Stanford NLP as Independent Evaluator in Dario's Oversight Plan

stanfordnlp · x · 2026-09-18

Stanford professor Christopher Manning responded to Dario Amodei's three-step plan by proposing universities—specifically Stanford NLP—as independent third-party evaluators of frontier AI, arguing universities are better suited than any other organization for key parts of the work. Researcher jchudnov co-signed with an inside view of Stanford's credentials: Terminal Bench Science (real agentic tasks scientists care about, built with Steven Dillmann), Program Bench (co-developed with a Stanford PhD who co-created SWE-bench), and open datasets/training efforts like DataComp, OpenThoughts, and Percy Liang's fully open-source Marin project.

Original post →

More from Safety

Safety channel →