GPT-6 Astra Scores 100% on ExploitBench as OpenAI Blocks PoC Exploit Requests

OpenAI on Thursday officially unveiled GPT‑6 Astra, which it described as the "world's most intelligent and aligned model." The development comes days after the artificial intelligence (AI) company said the model had reached the "Critical" cybersecurity capability threshold under its Preparedness Framework. "Astra is state-of-the-art on computer use, browsing, software engineering,

Fact: OpenAI has unveiled GPT-6 Astra, describing it as highly intelligent and aligned, and noting that it has reached a 'Critical' cybersecurity capability threshold under its Preparedness Framework and achieved 100% on ExploitBench, while blocking proof-of-concept exploit requests. Fact: These details indicate the model is both extremely capable in offensive security reasoning and subject to tighter controls on exploit generation. RealGround analysis: Such a capable model heightens the risk of malicious AI use if safety controls are bypassed, misconfigured, or inconsistently applied across products and integrations. RealGround analysis: Organizations deploying or integrating similar frontier models should implement continuous AI red teaming, secure agent design, and readiness assessments to validate that exploit-blocking and usage policies remain effective against evolving attack techniques.