OpenAI will train on Daybreak Blue session data starting November 29
OpenAI notified Trusted Access program members that beginning November 29 it will use Daybreak Blue cybersecurity session data to train its models, including when users have disabled training. Reporting attributed to The Decoder describes an initial email confirming the shift and a second message after backlash that promised a separate opt-out. The policy involves session data that was previously treated as off-limits for training, and the coverage links the move to renewed scrutiny after mathematicians suspected OpenAI had used unpublished research without consent, though those allegations are not confirmed in the available excerpts. Operators weighing sensitive workloads should note that familiar training toggles may not govern this session class until the promised opt-out is defined and implemented.
OpenAI will train on Daybreak Blue session data starting November 29
Starting November 29, OpenAI will use data from Daybreak Blue sessions to train its models, even when users have turned off training. An email to Trusted Access program members confirmed the change, and OpenAI followed up with a second email promising a separate opt-out after backlash.
Key takeaway
Do not assume a global training-off setting keeps Daybreak Blue session data out of model improvement after November 29.
What happened
According to The Decoder and descriptions of program email in the evidence packet, OpenAI told Trusted Access members that starting November 29 it will train on Daybreak Blue session data even when users have turned off training.
After backlash, OpenAI sent a follow-up email promising a separate opt-out; reporting also notes the change revives concerns from a prior dispute with mathematicians over suspected use of unpublished research without consent, without confirming those allegations in the excerpt.
Evidence
Training on Daybreak Blue sessions begins November 29 even when users disable training.
The Decoder · attributed
Starting November 29, OpenAI will use data from Daybreak Blue sessions to train its models, even when users have turned off training.
Trusted Access members were notified by email, with a second email promising opt-out after backlash.
The Decoder · attributed
An email to Trusted Access program members confirmed the change, and OpenAI followed up with a second email promising a separate opt-out after backlash.
Coverage ties the policy to prior consent concerns involving mathematicians and unpublished research.
The Decoder · attributed
The move revives concerns from the dispute with mathematicians who suspected OpenAI used their unpublished research for training without consent.
Why it matters
Builders and researchers must treat product privacy toggles as potentially non-binding for designated programs before sharing cybersecurity or unpublished material.
Limits and uncertainties
The excerpt does not establish the scope of Daybreak Blue data covered, the exact separate opt-out mechanism, or whether the policy applies beyond Daybreak Blue.
Mathematicians' allegations about unpublished research are referenced but not confirmed in the available reporting.
Coverage is based on described emails rather than demonstrated training outcomes.
Practical implications
Verify opt-out procedures for Trusted Access and Daybreak Blue before submitting sensitive session content.
Document which OpenAI surfaces and program tiers may override standard training-off settings in your risk reviews.
What to watch
Publication of the promised separate opt-out process and whether it applies before or only after November 29.
Trusted Access member communications clarifying enforcement relative to global training controls.
