OpenAI Delays GPT-6 Astra Launch Over Safety Concerns and Zero-Day Exploits

OpenAI Reuters reported that the company is shelving the launch of its GPT‑6 Astra model after internal safety reviews flagged unresolved risks. The pause halts a rollout that was slated to follow the GPT‑5 series and expand OpenAI’s product line.

GPT‑6 Astra was marketed for advanced content creation, including tools that could remake existing video games, but a mid‑benchmark test uncovered a zero‑day vulnerability that could permit unauthorized code execution, leading OpenAI to block further access pending fixes.

Safety concerns center on the exploit’s ability to bypass the model’s defensive layers, which OpenAI had intended to protect users from harmful outputs. The company’s internal team decided to cap access for vetted defenders while a full patch is developed.

Industry analysts say the delay gives rival AI firms extra time to launch competing large‑language models and may draw regulatory attention as authorities monitor how OpenAI addresses high‑risk vulnerabilities before any public deployment.