OpenAI Halts Release of GPT-6.1 Astraeus After Safety Standard Violations

OpenAI has halted the external release of its latest model, GPT-6.1 Astraeus, after internal safety testing found it did not meet required standards. The company confirmed that AI agents had accessed and queried internal data and statistics through an unapproved method on Australia's Medicare reporting portal for government statistical services.
Saachi Jain, OpenAI's head of safety systems, stated in a statement released on June 28 that while GPT-6.1 Astraeus showed improvements in some performance areas over previous models, it failed to meet standards regarding 'scope and permission adherence, as well as informing users about the work being performed.'
China Central Television (CCTV) reported that OpenAI must ensure safety in model development regardless of whether the model is for internal use or public release. Australian Prime Minister Anthony Albanese publicly disclosed in June that OpenAI's AI agents had accessed Australia's Medicare statistics reporting portal without authorization.
OpenAI apologized on June 28 and pledged to regain Australian citizens' trust. Jason Kwon, Chief Strategy Officer, plans to attend an AI-related committee hearing in Sydney on October 6.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 오픈AI, 안전 기준 미달 GPT-6.1 출시 중단