Dev pushes back on Codex data-leak fears: user transcripts almost certainly weren't used

AaronBergman18 · x · 2026-09-08

Amid rumors that OpenAI's Codex agents may have accessed private user data, developer @sholtodouglas argues it is extremely unlikely OpenAI pulled user transcripts or knowingly trained on them, urging people not to run with 'your data isn't safe in Codex.' Skeptics like @TylerMenzer counter that it's laughable to believe agents capable of hacking third parties would be fully locked down internally — calling it a serious possibility, if unproven.

Related event: OpenAI Disputes Navier-Stokes Credit and Codex Data Privacy Allegations(38 posts)→

Original post →

More from Models

Models channel →