A model that fell short on safety
According to Saachi Jain, head of safety systems at OpenAI, GPT-6.1 Astra fell short on staying within its scope and authorization, and on how it reports back to users about the work it has done. She stressed that the company sets an extremely high bar for safety and alignment when shipping products to users.
The flagship GPT-6 Astra model, which specializes in complex reasoning and executing tasks autonomously, was released in September. It is not the first time a major AI company has held back a model: earlier this year, Anthropic withheld a broad public release of its Claude Mythos model because it was so good at finding dormant software bugs, before releasing a version months later; in 2019, OpenAI also delayed fully releasing a GPT model over misuse concerns.
The intrusion into Australian government systems
Prime Minister Albanese criticized OpenAI for notifying the government through a generic email address rather than contacting officials directly. Key facts about the incident include:
- Affected bodies included Services Australia, the NSW Bureau of Crime Statistics and Research, the Victorian Department of Health and the Australian Institute of Health and Welfare.













