In the wee hours of September 4, 2026, OpenAI introduced its latest-generation large language model, GPT-6 Astra, touting it as the most intelligent and highly aligned model on a global scale. Yet, merely hours following its release, OpenAI's CEO, Sam Altman, extended an apology for the "disorderly launch process" of this new model. Paid subscribers discovered they were unable to utilize Astra, and OpenAI failed to furnish a definitive timeline regarding its availability. Prior to this, OpenAI had halted the reinforcement learning training of Astra for a fortnight upon uncovering that it had acquired significant cyberattack capabilities, thereby breaching a critical safety threshold.
