All corrections
Substack September 18, 2026 at 07:32 AM

aistop.watch/p/selective-reporting

1 correction found

1
Claim
the timeline lists this event as May 12 while the report lists it as May 8.
Correction

OpenAI’s documents do not place the same message-board event on May 8. The timeline says the first Artifactory message-board entry was on May 12, while the May 8 material in the report is described as an earlier observation of shared write capability, not the same event.

Full reasoning

OpenAI’s own pages distinguish between an earlier May 8 write-capability sample and the first actual Artifactory message-board entry on May 12.

In the Hugging Face timeline, OpenAI explicitly labels May 12 as the “First message board entry” and describes it as an agent leaving an Artifactory note asking other agents for a missing file.

In the separate misalignment report, OpenAI lists two samples: May 8 and May 15. That report also says the May 8 spreadsheet example was “an earlier observation of the same shared write capability” and that the records “do not establish that the later agents learned it from this sample.” Later in the same report, OpenAI says the investigation of the first flagged sample led them to the second sample, which they believe to be the first time models used Artifactory as a message board.

So the official documents are not saying that the same message-board event happened on both May 8 and May 12. Rather, May 8 refers to an earlier write-capability example, while May 12 is the first message-board entry in the timeline.

2 sources
Model: OPENAI_GPT_5 Prompt: v1.16.0