OpenAI plans to use Daybreak Blue session data for training its models from 29 November, including when users have disabled training. After the negative reaction, a follow-up letter promised a separate opt-out from such use.
The information is published by The Decoder; the provided package contains only metadata and a brief summary of the publication, with no primary statement by OpenAI or independent confirmation. Therefore, policy details and their actual application require further verification.
The situation again raises the question of consent to use data for training. The Decoder links it to the prior dispute about mathematicians who suspected OpenAI of using unpublished research without permission; this remains a description of suspicions, not established fact.
editorial commentary
Why it matters
Likely consequence — stronger requirements for explicit consent and a separate opt-out from the use of user sessions. The next observed signal — an official clarification from OpenAI and a description of the opt-out mechanism. Substantial uncertainty remains: only a meta-description from The Decoder is available, with no primary confirmation or implementation details.