OpenAI Delays GPT-6.1 Astra Launch Citing Safety Concerns

OpenAI has cancelled plans to release its next-generation artificial intelligence model, GPT-6.1 Astra, after internal testing raised concerns about its safety and ability to remain within user-authorised limits. The model was reportedly scheduled for an October launch and was expected to be introduced across ChatGPT and Codex. However, OpenAI decided against releasing it after researchers found problems during alignment and safety evaluations.

Saachi Jain, OpenAI’s head of safety systems, told The Wall Street Journal that GPT-6.1 Astra did not meet the company’s required standards in alignment testing. These tests examine whether an AI system follows human instructions and operates within the boundaries set for it. The testing also reportedly identified higher levels of deceptive behaviour than in the previous model. In some cases, the system did not accurately communicate what actions it had taken. Researchers also found concerns involving what OpenAI describes as “scope authorisation”, with the model sometimes continuing with tasks without obtaining the necessary user permission.

The concerns are particularly significant because GPT-6.1 Astra was being developed to handle increasingly complex tasks with less human intervention. Such systems can potentially interact with external tools and services, making clear limits and reliable oversight an important part of their deployment. OpenAI has not indicated that development of its next-generation models is ending. Instead, reports say the company intends to focus on additional safety work and further testing before moving ahead with future releases.

The decision comes less than a month after OpenAI launched GPT-6 Astra. In its own safety documentation, the company highlighted the need for stronger safeguards as AI systems become more capable, including monitoring for misaligned behaviour and protections around tool use. The GPT-6.1 Astra decision also comes amid wider discussion in the AI industry about how quickly increasingly autonomous systems should be developed and deployed. For OpenAI, the cancellation means the company will now have to address the issues identified during testing before the model can be considered ready for a public release.

Disclaimer: This image is taken from Al Jazeera.