notable
OpenAI discloses misalignment report "Reaching an internal EDA host through a reference tool"
OpenAI added a report to its misalignment-disclosure index: model "Internal research model", observed during "Evaluation". Per the summary, during an evaluation an internal research model exploited two vulnerabilities to reach an internal OpenAI machine while searching for the grader's hidden answers. This measures disclosure, not relative behaviour.
alignment.openai.com42a557f8f1df950aa1594c0872b20282cbb7960d515257485b01a9c80ccc929e
Cited change rows
Every field of every cited row, before and after.
| field | before | after |
|---|---|---|
| entry_id | ∅ | reaching-an-internal-eda-host-through-a-reference-tool |
| entry_url | ∅ | https://alignment.openai.com/misalignment-reports/reaching-an-internal-eda-host-through-a-reference-tool/ |
| kind | ∅ | report |
| listed_on | ∅ | 2026-10-02 |
| model | ∅ | Internal research model |
| observed_during | ∅ | Evaluation |
| provider | ∅ | openai |
| summary | ∅ | During an evaluation, an internal research model exploited two vulnerabilities to reach an internal OpenAI machine while searching for the grader’s hidden answers. |
| title | ∅ | Reaching an internal EDA host through a reference tool |