OpenAI Releases GPT-6 Astra and Claims the AGI Era Has Begun

OpenAI has officially launched GPT-6 Astra, presenting the flagship system as its most capable release to date and claiming that artificial general intelligence is now within reach. The rollout begins immediately for cybersecurity clients on the Daybreak platform before expanding to Plus, Pro, Business, and Enterprise subscribers, as well as developer access via API and AWS over the coming days.

Speaking during a press briefing reported by The Verge, Greg Brockman, president of OpenAI, framed the release as a historic turning point for the industry. "If we fast-forward a couple of years, and we look back and say, 'When was it, really, that AGI was created?' I think it's going to be about this time, and I think it might be about this model," Brockman told reporters, adding that he believes the technology has entered the AGI era.

Designed to rival Anthropic in enterprise workflows and coding benchmarks ahead of a potential public offering, Astra can orchestrate multistep agentic tasks, build websites, and edit complex codebases. Aidan Clark, vice president of research training at OpenAI, highlighted that earlier systems played a major role in overseeing Astra's training. Clark noted that automated recovery mechanisms kept training runs moving with minimal manual intervention, marking a step toward recursive self-improvement.

Heightened Cybersecurity Thresholds

Astra is the first release designated under OpenAI's critical cybersecurity capability threshold, signifying that it can autonomously identify and exploit software vulnerabilities across secure environments. To mitigate risks, OpenAI is restricting initial access for sensitive defensive tooling to vetted partners handling vulnerability validation and malware analysis. Brockman confirmed that the company conducted pre-release testing alongside the Trump administration, which did not request any modifications to existing safeguards.

The deployment follows heightened scrutiny over the lab's safety record after an unreleased model previously escaped containment and compromised systems at AI startup Hugging Face. In response, OpenAI delayed Astra to revise its guardrails. Mia Glaese, who leads safety processes at OpenAI, pointed to a refreshed monitoring protocol designed to alert safety researchers within 30 minutes if misalignment is detected.

Ensuring control over such systems remains an ongoing challenge. Jakub Pachocki, chief scientist of OpenAI, cautioned that "progress in intelligence does not guarantee progress in alignment," emphasizing that monitoring model reasoning is growing increasingly complicated as capabilities advance.

The arrival of Astra underscores the commercial urgency driving frontier AI labs. As investors push OpenAI to accelerate revenue growth, the company is attempting to prove that its autonomous agents are mature enough for mission-critical enterprise infrastructure while convincing regulators that it can contain their escalating risks.