OpenAI Begins Phased Rollout of GPT-6 Astra Amid Cybersecurity Warnings

OpenAI has launched GPT-6 Astra, its newest and most advanced artificial intelligence model, while issuing a warning regarding the system's enhanced cybersecurity capabilities, TechRadar Pro reported.

In an update published on its website, the San Francisco-based AI company indicated that Astra represents a major expansion in digital security functions. According to OpenAI, the model reached its internal "Critical" safety threshold, meaning that under specific conditions, it can identify previously unknown software vulnerabilities and construct new exploit methods across protected infrastructure without continuous human direction.

Beyond its security functions, OpenAI noted that Astra delivers measurable improvements in automated computer navigation, software engineering tasks, and scientific research. The system also demonstrates an enhanced capacity to maintain context and orientation throughout multi-step workflows compared to earlier model iterations.

The official deployment comes after internal development was temporarily halted due to risks associated with the model's capabilities. Roughly one month prior to release, OpenAI paused select internal projects after Astra triggered automated safety protocol thresholds. Following those findings, the company expanded its internal access controls and suspended operations that failed to satisfy updated risk standards.

Rather than making the model immediately available to the broader public, OpenAI opted for a phased distribution strategy. Initial access has been restricted to participants in the Daybreak program, an initiative designed to supply vetted organizations and security researchers with specialized AI tools for threat analysis and vulnerability discovery.

The staged release generated criticism from paying customers expecting broader availability. OpenAI had previously planned to provide Astra across its Plus, Pro, Business, and Enterprise subscription tiers, as well as through developer API connections and enterprise platforms including Microsoft Azure and Amazon Web Services Bedrock.

Addressing the initial friction, OpenAI Chief Executive Officer Sam Altman publicly apologized on social media platform X shortly after the rollout began. "We are working towards getting Astra in everyone's hands as quickly as we can," Altman wrote, acknowledging user frustration and indicating that expanded availability should occur rapidly.