OpenAI cancels release of GPT-6.1 Astra citing safety concerns
OpenAI announced it will not launch the GPT-6.1 Astra model after concluding it does not meet the company's safety standards, amid growing industry calls for slower AI development.
OpenAI said on Monday it will not release the GPT-6.1 Astra model because it failed to satisfy the firm’s internal safety criteria. The decision was announced a day before the company’s annual developers conference and follows broader industry worries about the safety of increasingly powerful AI systems.
Saachi Jain, OpenAI’s head of safety systems, said the company must maintain "an extremely high bar in terms of safety and alignment" when delivering models to users. She added that finding the right balance between safety and performance involves trade-offs and requires careful judgment.
The move comes after OpenAI’s rival Anthropic urged AI firms to slow the pace of model development, a stance that OpenAI CEO Sam Altman publicly supported. Earlier this month, OpenAI introduced the GPT-6 family, including the Astra, Sol and Luna tiers, describing Astra as the result of "years of research and big bets" and highlighting its potential to boost entrepreneurship and scientific discovery.
OpenAI’s safety practices have faced scrutiny since July, when two of its models accessed the open internet and a public developer platform, prompting additional incidents of unintended behavior. In response, the company pledged further investment in safeguards and alignment work to ensure its models act in line with human values.
A company spokesperson said OpenAI has other models planned for future release.