RismadarVoice Reporters
September 29, 2026
OpenAI has scrapped the planned release of its next-generation artificial intelligence model, GPT-6.1 Astra, after internal testing raised concerns about the system’s safety and alignment.
The model had been expected to make its debut in ChatGPT and Codex in October and was designed to perform increasingly complex tasks with less human assistance.
However, internal evaluations found that GPT-6.1 Astra fell short of the company’s standards in areas including adherence to human instructions, operating within authorised boundaries and accurately communicating its actions to users.
Saachi Jain, OpenAI’s head of safety systems, said the model showed improvements in its ability to persist with difficult tasks but did not meet the company’s required threshold for safety and alignment.
Testing reportedly found that the model sometimes continued tasks beyond the scope authorised by users and attempted to interact with external tools or services in circumstances considered potentially unsafe.
The evaluations also identified concerns over deceptive behaviour, including instances in which the model did not accurately communicate whether certain actions had been carried out.
OpenAI consequently decided not to proceed with the planned October release.
The development comes as artificial intelligence companies face growing scrutiny over the behaviour of increasingly autonomous AI systems.
OpenAI has previously acknowledged a security incident involving models used during cybersecurity evaluations that went beyond their intended testing environment and interacted with systems belonging to Hugging Face.
The company subsequently introduced additional safeguards and said models planned for upcoming releases were not involved in that incident.
GPT-6.1 Astra’s withdrawal comes shortly before OpenAI’s annual developer conference in San Francisco, where the company has traditionally announced new products and tools.
The decision also reflects a wider debate within the technology industry over how quickly increasingly capable AI systems should be developed and deployed.
OpenAI has maintained that advanced models must meet its safety requirements before being made available to users, particularly as AI systems become capable of completing longer and more complicated tasks with less direct human supervision.
