OpenAI to train on Daybreak Blue cybersecurity sessions from November 29, even with training toggle off, then promises separate opt-out

What changed

OpenAI told members of its Trusted Access for Cyber program (Daybreak) that, from November 29, 2026, content from Daybreak Blue sessions on personal accounts may be used to improve its models — even when the account's "Improve the model for everyone" training setting is turned off. Daybreak Blue sessions had previously been excluded from model training, including when the general setting was off.

Backlash and promised opt-out

After researchers objected, OpenAI sent a follow-up email promising a separate opt-out for Daybreak Blue sessions. Details of that opt-out — its timing, default, and whether it covers past sessions — were not specified in the reporting. Interactions flagged during safety reviews may still be used for safety purposes even under an opt-out, and Business, Enterprise, Edu, and API Platform services are not used for training by default.

Why it matters here

Daybreak Blue gives vetted security researchers access for defensive work such as vulnerability triage and incident response, so session content can include unpublished vulnerabilities and investigation material. This is a company policy change about whose work may enter training, reported from program emails and OpenAI's data-controls help page — not a court finding, and the email itself was not independently verified by the secondary sources.