Yoav Goldberg: OpenAI's "cannot rule out" user data use is a lawyered admission

MannyKayy · x · 2026-09-09

Context: OpenAI congratulated mathematicians Levent Alpöge and Tristan Buckmaster, stating neither researchers nor agents saw their work before public release and no specific user data was accessed — but adding they "cannot rule out" that de-identified data from the mathematicians' product usage helped improve models used in the effort.

NLP researcher Yoav Goldberg argues OpenAI could definitively check: the dates are known, and they have the pipelines and logs. Since every OpenAI statement is PR- and lawyer-vetted, "cannot rule out" can realistically only mean that recent user data did flow into a pipeline behind a model used here.

Related event: OpenAI's admission it can't rule out training on anonymized user data sparks research-priority controversy(9 posts)→

Original post →

More from Models

Models channel →