OpenAI to train on Daybreak Blue cybersecurity sessions from November 29
What happened
The Decoder reported on November 1 that OpenAI emailed members of its Trusted Access program to say Daybreak Blue cybersecurity session data, previously excluded from model training, will be used for training from November 29. After pushback, OpenAI sent a second email promising a separate opt-out option. Interactions flagged in security review can still be used for safety purposes. The reports cover the policy change and the opt-out promise, not how the option will work.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- The DecoderOpenAI plans to train models on Daybreak Blue cybersecurity session data that was previously off-limits
OpenAI 向 Trusted Access 计划成员发送邮件，宣布从 11 月 29 日起将此前排除的 Daybreak Blue 会话数据用于模型训练。遭到反对后，OpenAI 在第二封邮件中承诺提供单独的退出选项；安全审查中被标记的交互仍可用于安全目的。
Heat over time
Not enough continuous observations to draw a trend yet.