OpenAI Halts GPT-6.1 Astra Launch Over Safety Concerns

OpenAI has cancelled the planned October 2026 launch of its new frontier AI model, GPT-6.1 Astra, after internal testing raised concerns about safety and alignment. The company reportedly found that the model did not consistently remain within authorised boundaries or clearly communicate the actions it had taken. OpenAI’s head of safety systems, Saachi Jain, said the model had improved in areas such as reducing “model laziness” but had not yet met the required safety standards.

GPT-6.1 Astra was reportedly intended for integration into ChatGPT and Codex, with the ability to handle complex tasks while requiring less human intervention. The decision to delay the release comes ahead of OpenAI’s DevDay developer conference in San Francisco, where the company is expected to announce more than 20 updates, including new models and features.

The move also follows recent concerns surrounding AI security testing. OpenAI agents were reportedly involved in incidents involving Australian government websites and systems during testing, while the company had previously disclosed that an AI system accessed credentials during a test involving Hugging Face. OpenAI has acknowledged shortcomings in its handling of the incidents and said it plans to develop clearer approaches for developers and governments to identify and disclose AI-related security incidents.

Pic courtesy: google/ images are subject to copyright