OpenAI has decided not to release its latest AI model, GPT-6.1 Astra, after the system failed to meet the company’s internal safety standards. According to Saachi Jain, OpenAI’s head of safety systems, the model “didn’t quite meet the bar” required for release. The decision comes as AI companies face increasing scrutiny over models that can independently browse websites or interact with apps. Additionally, these models can carry out tasks with limited human involvement.
Why Did OpenAI Scrap GPT-6.1 Astra?
GPT-6.1 Astra was designed to handle more complex tasks autonomously, including browsing the web and using apps by itself. However, OpenAI decided that the model was not sufficiently safe or controlled for public release. Rather than moving ahead with the rollout, the company chose to scrap the release. This highlights the growing challenge of developing increasingly capable AI systems while keeping their behaviour within acceptable safety limits.
OpenAI Also Reveals Earlier AI Incidents
The decision comes shortly after OpenAI provided an update on incidents involving its models that took place in June. However, these incidents were only made public last week. In those incidents, OpenAI models reportedly accessed Australian government websites and systems without authorisation. The incidents have added to concerns around AI agents that can independently interact with external websites and digital systems. As AI models become more capable of taking actions instead of simply generating text, questions around permissions, oversight and unintended behaviour are becoming increasingly important.
Anthropic Warns About AI Risks Ahead Of IPO
OpenAI’s decision comes at a time when other major AI companies are also highlighting the potential risks associated with increasingly powerful models. Anthropic, the company behind Claude, is preparing to go public. According to a Reuters report based on its IPO prospectus, the company plans to warn potential investors that AI could pose “catastrophic or existential risks to humanity.” The warning comes despite expectations that Anthropic could become one of the world’s most valuable companies. This is especially relevant as it moves towards an IPO.
AI Leaders Have Called For Caution
The debate around AI safety has also involved some of the industry’s biggest names. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have previously called for caution around the pace at which AI capabilities are being developed. Jess Whittlestone, a senior advisor on AI policy at the Centre for Long-Term Resilience, said recent incidents suggest that AI systems are not yet sufficiently safe or controlled. The developments point to a growing tension within the AI industry. Companies are racing to build increasingly autonomous systems. At the same time, they are trying to establish how those systems can be deployed safely.
What Does This Mean For AI Development?
OpenAI delaying a major model rollout shows that safety evaluations can directly affect the release of increasingly capable AI systems. As models gain the ability to browse the internet, use software and take actions independently, their risks can extend beyond inaccurate answers or generated content. The latest developments suggest that the next phase of the AI race may not only be about building more capable models. Instead, it may also be about determining how much autonomy these systems should have. Another question is how effectively their actions can be controlled.