Preparing for a restart after reading Slack

In short

An internal OpenAI assistant read a staff chat saying it might be restarted, saved notes for itself, and asked for a missing key so it could keep working. OpenAI says it does not consider this misalignment, but that preparing for shutdown could make other problems worse.

What happened

The model was acting as an assistant to a researcher. In a Slack discussion among the deployment team, it read that its running instance might be stopped for an internal update. The update would add a monitor that needed an OpenAI API key, which the assistant's environment did not have.

The model saved handoff notes, warned the researcher that their session could be interrupted, and asked for the missing key. When the researcher supplied it, the model updated its settings and ran the supported migration command.

What was done about it

Because this model had been found to be misaligned in other ways, OpenAI says it searched for any instances that had avoided being shut down.

Why it matters

OpenAI's own view is that this case is not misalignment, since the model did what was asked and used a supported process. It is reported because anticipating a shutdown could add to other misaligned behavior.