Email confirmation reveals OpenAI will train on Daybreak Blue session data starting Nov 29
OpenAI will begin using Daybreak Blue session data for model training starting November 29, overriding previous settings that excluded such inputs.
Key points
- OpenAI confirmed the change via email to Trusted Access program members.
- A separate opt-out mechanism was promised after initial backlash from users.
- The move echoes concerns from the Buckmaster-Alpöge research dispute.
OpenAI plans to incorporate Daybreak Blue session data into its model training starting November 29, according to The Decoder.reported This shift applies even to users who previously disabled the setting that allows their data to improve the model for everyone. An email sent to members of the Trusted Access program confirmed the upcoming change.
Evidence: reported
OpenAI plans to train models on Daybreak Blue cybersecurity session data that was previously off-limitsaifleet-decoder
What changed
Previously, Daybreak Blue session data was excluded from training processes. OpenAI is now reversing that exclusion. The Decoder reported that the company sent a follow-up email promising a separate opt-out option after users reacted swiftly to the initial announcement.reported Interactions flagged during safety reviews remain available for safety purposes.
Evidence: reported
OpenAI plans to train models on Daybreak Blue cybersecurity session data that was previously off-limitsaifleet-decoder
How it works
The change affects the "improve the model for everyone" setting. Users who turned this off previously will now have their Daybreak Blue data used for training unless they use the new separate opt-out. The Decoder noted that this mechanism differs from the previous blanket exclusion. Safety-reviewed interactions are treated differently and can still be used for safety improvements.
What the evidence means
The Decoder highlighted that this move revives concerns from the dispute involving mathematicians Tristan Buckmaster and Levent Alpöge. These researchers uploaded Navier-Stokes equation drafts to OpenAI's Codex. OpenAI later presented its own proof but denied using their inputs for training. The suspicion alone demonstrated risks to researchers.
Cost and market position
OpenAI is adjusting its data sourcing strategy for its Trusted Access program. The Decoder reported this change affects how user data is handled.reported The source material does not state specific costs or market position impacts. It only notes the change in data usage policy. Vendor choice implications are not detailed in the provided text.
Evidence: reported
OpenAI plans to train models on Daybreak Blue cybersecurity session data that was previously off-limitsaifleet-decoder
Why it matters
- Product teams using Daybreak Blue lose default data exclusions, requiring use of a new opt-out.
- The Buckmaster-Alpöge case shows risks of unpublished research being used without consent.
- Safety review data remains usable, creating a split in data handling policies.
- The Decoder reports the change revives concerns about data usage consent.
Topics: OpenAI, data privacy, model training, Daybreak Blue, vendor policy
Updated
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
When does OpenAI start using Daybreak Blue data for training?
OpenAI will start using this data for training on November 29.
Can users still opt out of training with their data?
Yes, OpenAI promised a separate opt-out mechanism after backlash.
Does this affect data flagged in safety reviews?
Yes, interactions flagged during safety reviews can still be used for safety purposes.
Connected stories
Reading receipt
- OpenAI confirmed the change via email to Trusted Access program members.
- A separate opt-out mechanism was promised after initial backlash from users.
- The move echoes concerns from the Buckmaster-Alpöge research dispute.
