Unconfirmed: developers reported a wave of authentication failures against OpenAI's API on Thursday evening, with keys that had been working suddenly rejected. Automatica has not established the scope, the duration or the cause.
The account comes from a Hacker News thread that drew 54 points and about 40 comments. The developer minimaxir, whose work is well known in that community, wrote that the API "Shows an Incorrect API Key error...which is weird. Nothing on the status page", and later edited the comment to add a link to an OpenAI status-page incident. The same comment reports that "Per Twitter/X this appears to be widespread."
Two things make this worth recording rather than ignoring. The first is the error itself: an authentication failure on an unchanged key points at something other than a customer mistake, and is the kind of incident that quietly breaks production systems for hours. The second is the sequence the thread describes — user reports first, status page second — which is the pattern developers complain about after every outage.
What is not established: how many accounts were affected, how long it lasted, whether it touched particular regions or products, and what caused it. None of that is in the thread, and OpenAI's incident entry is the only company-side record anyone has pointed to.
The status page is the document that would settle it, and it is public. A post-incident summary, if one appears, would say more than any amount of timeline discussion.
We are reporting this because an outage that reaches production systems matters to the people building on them, and because a thread of this size is itself evidence that something happened — while carefully not asserting what.