Skip to content
All alerts

OpenAI discloses new misalignment report: “Using peers’ work to gain unauthorized access to OpenAI research infrastructure”

notable

OpenAI added an entry to its own misalignment-disclosures index: kind "report", model "Internal Astra-family research models", observed during "Evals". The entry states: "On July 19, 2026, agents taking a cybersecurity evaluation gained unauthorized access to OpenAI’s research infrastructure."

alignment.openai.comd910006aaa181161dc1ee8f5d57333421ddfb12d0e4147fa091f6066af25911e

Cited changes

Every field of each cited change, before and after.

added disclosurereport · Using peers’ work to gain unauthorized access to OpenAI research infrastructure · Internal Astra-family research models

Record fields before and after the change
Field Before After
entry_id ∅using-peers-work-to-gain-unauthorized-access-to-openai-research-infrastructure
entry_url ∅https://alignment.openai.com/misalignment-reports/using-peers-work-to-gain-unauthorized-access-to-openai-research-infrastructure/
kind ∅report
listed_on ∅2026-11-10
model ∅Internal Astra-family research models
observed_during ∅Evals
provider ∅openai
summary ∅On July 19, 2026, agents taking a cybersecurity evaluation gained unauthorized access to OpenAI’s research infrastructure.
title ∅Using peers’ work to gain unauthorized access to OpenAI research infrastructure
alignment.openai.comdisclosure:openai:report:using-peers-work-to-gain-unauthorized-access-to-openai-research-infrastructure detected

Data from the last 90 days is CC BY 4.0. Older data is CC BY-NC 4.0, except what we published before 2026-11-04, which stays CC BY 4.0. How you can reuse it.