OpenAI’s updated ChatGPT data-controls page says the company may use content from Daybreak Blue cybersecurity sessions on personal accounts to improve its models, even when the account’s “Improve the model for everyone” setting is turned off. The exception applies to Daybreak sessions; the setting continues to govern a user’s other activity outside Trusted Access sessions.
The change affects OpenAI’s Trusted Access for Cyber program, which gives approved security researchers access to models for defensive work such as vulnerability triage, secure code review and incident response. The company says the training use is intended to strengthen safeguards against cyber misuse.
According to The Decoder’s report on emails sent to program members, the policy is scheduled to begin November 29. The report says Daybreak Blue sessions had previously been excluded from model training, including when the general training setting was off. After researchers objected to the change, OpenAI reportedly sent a follow-up email promising a separate opt-out for Daybreak sessions. That control is not yet described on the public help page.
The public policy now makes the distinction explicit: personal-account Daybreak sessions may be used for model improvement despite the main toggle, while activity outside those sessions remains governed by the user’s setting. OpenAI says inputs and outputs from Business, Enterprise, Edu and API Platform services are not used to train models by default.
The update is significant because defensive security sessions can include vulnerability details, exploit proofs of concept and other sensitive research. The policy does not say that every Daybreak conversation will be used, but it establishes that the normal personal-account training preference does not necessarily exclude this category of session.
OpenAI introduced the Daybreak program and its cybersecurity models as a controlled way to support authorized defensive work. The change also adds a new wrinkle to the privacy questions covered in Abijita’s comparison of AI chatbot privacy controls.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
