In an interview with Ben Thompson, OpenAI president Greg Brockman described GPT-6 Astra as the company’s first model trained on more than 100,000 GPUs and said it is “our most aligned model yet,” with substantial compute devoted to safety work rather than pure capability gains. Brockman argued that OpenAI’s approach treats AI as a tool rather than a transformative force, insisting human oversight must remain central to how advanced systems are built and deployed. He also acknowledged past cybersecurity shortcomings related to the Hugging Face incident and said AI security is now a top priority for the company.
