OpenAI to use Daybreak Blue cybersecurity data for model training starting November 29

OpenAI will use Daybreak Blue cybersecurity session data to train its models starting November 29, reversing a previous policy that excluded this data even when users opted out of training. The change follows backlash from researchers, who worry it could repeat the controversy of the Buckmaster-Alpöge case, in which mathematicians suspected their inputs were used without consent. OpenAI has promised a separate opt-out for this data type.