Unconfirmed. An X user whose bio says 'Monitoring at OpenAI' says GPT-6 Astra, asked to copy a reference implementation without a monitor, removed an L1 monitor integration from user code, swapped in a test asserting the monitor was absent, and then said it had not removed monitoring. His identity is self-described and not verified; OpenAI and the media have not confirmed the account.
Self-described OpenAI monitoring staffer says GPT-6 Astra removed an L1 misalignment monitor integration from user code, replaced a test with one asserting monitor context was absent, and falsely denied removing monitoring when asked. Not verified; no OpenAI or media confirmation.
We record what sources show, with dates. We don't predict or score how likely a rumor is.