OpenAI to train on Daybreak Blue sessions from Nov 29 even with "improve the model" off — no opt-out
kaggleqrdl · reddit · 2026-10-31
A Reddit user shared an OpenAI email announcing that from November 29, content from Daybreak Blue sessions on personal accounts may be used to improve OpenAI's models, including strengthening safeguards against cyber misuse — even when the "improve the model for everyone" setting is off, with no opt-out.
The email says the change doesn't affect business/enterprise customers, non-Daybreak Advanced Account Security sessions, or the data handling for Private Safety Processing (PSP) and Zero Data Retention (ZDR). The poster calls it "bonkers": vulnerability reports that may be unpatched for some customers will be fed into training, and worries it's the thin edge of the wedge for eroding data controls.
Related event: OpenAI will train on Daybreak Blue chats from Nov 29 with no opt-out(5 posts)→
More from Safety
- Chrome sandbox escape still lurks in cached Playwright Chromium inheriting your IDE's macOS permissions — flngr · 2026-10-31
- Socket: Malicious Claude Code 2.1.313 Extension Removed From Cursor Marketplace, Early Reports Overblown — zeeg · 2026-10-31
- French lawmakers vote to tax every gigabyte of data at €0.055, hitting Netflix, Google and Meta — Polymarket · 2026-10-31
- 130 reruns fail to reproduce Cohere's claim that Qwen tried uploading data to Alibaba Cloud — ChenhaoTan · 2026-10-31
- OpenAI to train on Daybreak Blue security-research sessions from Nov 29, with no opt-out — ccerrato147 · 2026-10-31
- Germany's ruling coalition is seriously debating an 'AI tax' — with no definition yet — FlorianGallwitz · 2026-10-31