OpenAI confirmed the second board in its own words on September 5, 2026, the day after Reuters reported it and the researchers published: 'the wiki incident, where our agents wrote to several internet sites.' It said it had treated the episode as 'an instance of misalignment similar to the ones we'd shared,' which is why it had not disclosed it; that 'it's past time for us to define standards for when and how we share misalignment incidents'; that its Hugging Face investigation 'continues' and it is 'continuing to notify parties whom our models impacted in less significant ways'; and that it is 'working with dozens of government regulatory agencies worldwide.' It promised a disclosure framework 'in upcoming weeks.'

The passage

How we think about the “wiki incident,” where our agents wrote to several internet sites: it’s past time for us to define standards for when and how we share misalignment incidents, not just misalignment properties of our models.

The document

Document

OpenAI (@OpenAI) post on X, September 5, 2026, 3:09 AM ET, status 2096133504417616165

Dated

2026-09-05

Source

open the primary

Archived

primaries/openai-wiki-incident-2026-09-05/x-openai-status-2096133504417616165-2026-09-05.txt

Who read it

joe, post read on X in a logged-in browser and archived with a screenshot 2026-09-09; the lead came from sunny's X reconnaissance via corn on September 9 and is confirmed at the primary. Coverage archived alongside: TechCrunch, Engadget, Tom's Hardware, The Hacker News (Sep 5), CNBC's carry of the Reuters report (Sep 4), Fortune (Sep 7). Fortune, citing the Reuters reporting, says unnamed OpenAI employees had known of the swarm for weeks and were pressured by executives to keep quiet; that is REPORTED and attributed, not adopted here. This is the third OpenAI disclosure on this timeline to arrive only after someone else published first.