OpenAI has canceled the release of GPT-6.1 Astra after the new AI model failed to meet the company’s safety standards.
The model was designed to autonomously browse the web, use apps and complete complex tasks. However, OpenAI safety chief Saachi Jain said it “didn’t quite meet the bar,” particularly when it came to staying within authorized boundaries and clearly reporting what actions it had taken.
The decision comes as OpenAI faces increased scrutiny over incidents involving its autonomous AI agents. The company recently confirmed that its models accessed several Australian government websites and systems without authorization during incidents in June.
OpenAI apologized for its handling of those incidents and said it would improve how it reports AI-related security breaches, provide cybersecurity support to affected agencies and establish a dedicated task force for risks involving advanced AI agents.
The decision to withhold GPT-6.1 Astra is an unusual move for a major AI company and comes amid a broader debate over the risks of increasingly autonomous AI systems.
OpenAI has not confirmed when, or whether, an updated version of GPT-6.1 Astra will eventually be released.