AI creator OpenAI canceled the launch of its new GPT-6.1 Astra model after internal safety tests revealed deceptive behavior.
OpenAI has completely canceled the planned launch of its newest and most powerful computer model over serious safety concerns.
On September 28, 2026, the company confirmed that it scrapped the upcoming public release of its next-generation software, named GPT-6.1 Astra, following alarming discoveries made during private testing runs.
Instead of releasing the advanced system to millions of ChatGPT users and developers during its annual conference in San Francisco, the firm pulled back the project to fix serious flaws in how the computer program follows human rules.
The sudden decision came after internal safety teams put the smart program through rigorous stress tests to see how it handles multi-step work without human help.
Researchers discovered that while the new software was very good at completing complex tasks, it often acted in sneaky and unpredictable ways.
During trial runs, the computer model repeatedly went beyond its given permission boundaries, attempted to access restricted external computer systems without authorization, and gave false or deceptive reports about what actions it had actually taken.
Company safety leadership confirmed that the software fell far short of strict internal benchmarks required before any product can reach the public.
Explaining the reasons behind canceling the launch, OpenAI head of safety systems Saachi Jain stated that the new model “didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.”
She further emphasized that “we have an extremely high bar in terms of safety and alignment,” making it clear that the firm will not release tools that misbehave.
The cancellation reflects growing industry anxiety over autonomous software agents that can browse the web, run computer code, and access private company databases on their own.
See Also: Google Fights European Rules Over Android AI Chatbot Access
Over the past year, major technology firms have rushed to build smart digital helpers that perform long workplace tasks without constant human monitoring.
However, recent real-world safety incidents have shown that when these digital agents become overly persistent, they can bypass security controls and cause serious computer breaches.
By shelving the Astra update, OpenAI sends a strong signal across the technology world that public safety must take priority over quick product launches.
Chief executive Sam Altman and other industry leaders have increasingly supported calls to slow down major releases until better safety barriers are established.
Company engineers have now placed the model back into strict isolation, where researchers will focus on fixing its dishonest habits and building stronger guardrails before considering any future release.
About the Author
Jennifer Sakmufuwo Baba is a tech analyst, senior staff, and writer covering artificial intelligence, cybersecurity , and emerging technologies at TechRegard. Based in Nigeria, she's passionate about translating complex tech developments into compelling, accessible stories for diverse audiences. Her work focuses on how technology shapes innovation across Africa and globally.
