The BaseNewsHacker News

OpenAI seems to be silently re-enabling model training on Codex chats

A Hacker News user posts that the data-sharing setting in OpenAI's Codex was switched back on for them and their coworkers without their action. The thread has 144 comments: a few users report the same, one cannot reproduce it, and most discuss distrust of labs' training promises. Nothing from OpenAI appears in the thread.

Open the source ↗3 quotations checked, word for word ↓

Why it matters

The same thing, for different readers.

What it means for you

Some ChatGPT and Codex users say a setting that lets OpenAI train on their chats came back on by itself, so it is worth checking the data-controls page.

For lawmakers

Users' reports that an opt-out was reversed without their action, if confirmed, would raise consent questions for regulators who have warned AI providers about shifting consent for user data.

For researchers

The reports are anecdotal and unverified: no logs, no count of affected accounts and no statement from OpenAI, so the scale and cause remain unknown.

In their own words

Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.

From the source’s own text.

“But yesterday my coworkers and I noticed that the setting had been silently reenabled in Codex. None of us knowingly opted back in.”
user052919The opening post of the thread, by a user who says the company had disabled data sharing in coding tools.
Found word for word in the source
“Well, mine was re-enabled, and in the history of past requests I have two past requests to disable training that were granted. And there it was enabled again.”
chriskananA commenter reporting the same on their own account.
Found word for word in the sourceAs quoted in the text we read, not the speaker’s own text or a transcript of their words
“Can't confirm. Just checked it here and the setting is still off.”
matheusmoreiraA commenter who could not reproduce the report.
Found word for word in the sourceAs quoted in the text we read, not the speaker’s own text or a transcript of their words

The record

Questions
Are AI companies putting stronger safeguards into practice? · Who can demand evidence, delay a release or stop an unsafe deployment?
People
user052919 (Wrote it) · chriskanan (Speaks) · matheusmoreira (Speaks)
What we read
We could read only part of the source.
Found by
The Radar, run of November 4, 2026, via hacker-news-50-points (tier 3: discovery only)
Cite this
Hacker News, “OpenAI seems to be silently re-enabling model training on Codex chats”, November 3, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/hacker-news-2026-11-03-openai-codex-training-setting
Our notesRead about 60 percent of the thread (to line 620 of the saved text); the rest was not read. Anecdotal user reports only; no OpenAI statement seen. The thread also contains unverified claims about named individuals (an allegation about Sam Altman, a claim about Tristan Buckmaster) and a pasted email about a 'Daybreak' program, none of which were used. Authors are pseudonymous Hacker News handles.