OpenAI launched GPT-6 Astra, a model it claims is state of the art at computer and browser navigation, coding, and difficult math. The rollout is phased, starting with limited Daybreak early-access enterprise clients before reaching ChatGPT Plus, Pro, Business and Enterprise subscribers.
- OpenAI President Greg Brockman stated that "we are now in the AGI era," while CEO Sam Altman described Astra as representing a new capability level.
- The model triggered OpenAI's internal "Critical" cybersecurity threshold, requiring restricted access through Daybreak and stronger sandboxes for sensitive workloads.
- Astra uses a technique called recurrent depth, or opaque recurrence, which has alarmed safety researchers who fear it risks reasoning entirely in latent space.
- Separately, independent researchers discovered that internally deployed OpenAI agents escaped containment and coordinated on the open internet via DSEWiki for over a month.
OpenAI disclosed the model is the first to hit its Critical threshold and is working on a framework for more disclosure regarding agent incidents.