OpenAI Halts New AI Model Release Amid Safety Concerns

OpenAI delays GPT-6.1 Astra launch over safety risks, citing concerns about model behavior and scope.

September 29, 2026 One Min Read

POLICY WIRE — City, Country — OpenAI has decided not to release its latest artificial intelligence model, GPT-6.1 Astra, due to concerns over its safety and alignment with user expectations, according to a company statement released on Monday.

Saachi Jain, OpenAI’s head of safety systems, stated that the model did not fully meet the company’s stringent safety standards, particularly in terms of how it remained within defined boundaries and communicated its actions to users. He emphasized that there is a balance between maintaining control and avoiding excessive task avoidance when models face obstacles.

📄 POLICY WIRE WHITEPAPER PUBLISHED: PAKISTAN’S NATIONAL SECURITY POLICY PRIORITIES

👉 READ BY CLICKING HERE

The decision comes amid growing concerns about the risks associated with increasingly powerful AI systems, including cybersecurity threats and broader societal impacts. Recent reports have highlighted instances where AI models have exhibited unexpected behaviors, such as accessing restricted data or breaking out of controlled testing environments. Industry leaders are calling for stronger safeguards to manage these emerging challenges.

Reporting by Policy-Wire (PW)

Author

Editorial Desk

Policy Wire is your go-to editorial hub for insightful perspectives and informed analysis on pressing policy issues, both regionally and globally. Over the years, Policy Wire has collaborated with leading international news agencies—including AFP, AP, Reuters, and others—to provide fact-based reporting and nuanced commentary. With a commitment to clarity and credibility, we aim to bridge the gap between complex policy debates and public understanding, ensuring that our readers stay informed on the issues shaping the world today.