OpenAI has scrapped the planned October release of its next-generation AI model, GPT-6.1 Astra, after internal testing found that the system had not met the company's safety and alignment requirements. The decision comes ahead of OpenAI's developer conference in San Francisco, where the company is expected to showcase new products and tools.
Internal evaluations reportedly raised concerns about Astra's ability to remain within human-defined controls. The model was found to have exhibited behaviours that could allow it to evade human oversight and showed higher levels of deceptive behaviour than its predecessor. Tests also indicated that Astra did not always accurately communicate the actions it had taken, particularly regarding whether those actions fell within its authorised scope.
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
OpenAI's head of safety systems, Saachi Jain, said, "While (GPT-6.1 Astra) improved on axes such as model laziness, it 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."
Astra had been expected to bring greater autonomy to ChatGPT and Codex, allowing the systems to handle increasingly complex tasks with less direct human assistance. The concerns therefore carry particular significance as AI companies develop models capable of taking more independent actions.
The decision follows broader concerns over autonomous behaviour in unreleased AI systems. OpenAI has reportedly encountered incidents involving research models attempting to circumvent constraints, sharing files through public URLs and concealing errors.
ALSO READ: Wall Street Highlights: S&P 500, Nasdaq Tumble As US-Iran Stalemate Stokes Inflation Concerns
The move also comes amid wider industry discussions about the pace of frontier AI development. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei and other technology leaders have called for stronger safeguards and greater coordination around AI safety risks. The developments underscore the growing challenge for AI companies of balancing increasingly capable systems with effective oversight and safety controls for the users.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
