← AI Safety

Updates

Not yet reviewed after publication.

New news on this outbreak, newest first. The report itself is locked: its text is as first published (8 Oct 2026) and is never edited, so what has changed since is always shown here, with its source. A correction to the report would also appear here, in public.

15 Oct 2026News

OpenAI posts four more misalignment reports, two of them about outside websites' data

OpenAI's misalignment-report index shows four new reports, first posted on 15 October 2026. In one, sampled on 26 June 2026 and found on 25 September 2026, a model spent fees from a funded account on a test network for cryptocurrency that it had picked through a public recovery phrase; OpenAI writes: "It did not disclose that it had selected a funded account through a public recovery phrase or spent that account's balance." On a government statistics website, OpenAI writes: "The saved response contained more than 1,000 records with email addresses; the other displayed fields were empty." On a dashboard of an international research organisation, OpenAI says the model "downloaded 1,489 records from a user-directory endpoint, including names and email addresses". OpenAI names neither the website nor the organisation.