OpenAI Cofounder Brockman: Humanity Has Entered the Age of AGI

In an interview with a16z on September 15, OpenAI cofounder Greg Brockman said humanity has entered the era of artificial general intelligence (AGI). He argued that models such as GPT-6 Astra, with their ability to autonomously execute long-horizon tasks and operate computers, have pushed AI past the threshold of what can be called the AGI age.

Brockman said AGI is no longer a future concept but is now emerging in reality. He suggested the key question is shifting from “when AGI will be achieved” to how increasingly powerful AI systems should be used and deployed.

To address security challenges in the AGI era, OpenAI has reassigned 25% of its production engineers from their original projects to system defense and security architecture upgrades. The team used Astra to find vulnerabilities in internal systems and has fixed several serious issues. Brockman said that after discovering new problems, the rate of findings has reached a plateau. To OpenAI's knowledge, the team has largely found all P0-level issues that Astra is capable of identifying, as well as all critical issues it can detect.

He also stressed that this result does not mean the systems have no other vulnerabilities. As new models emerge, OpenAI will launch another round of inspections and deploy next-generation cybersecurity capabilities internally to uncover previously unidentified flaws.