OpenAI Halts New AI Model Release Due to Safety Concerns

OpenAI has delayed its GPT-6.1 Astra model after internal safety testing revealed safety issues. The ChatGPT-maker confirmed on Monday that the model didn't quite meet the bar for safety standards.

Saachi Jain, OpenAI’s head of safety systems, stated that while GPT-6.1 Astra improved on model laziness, it did not meet the bar in terms of staying within scope and authorization and how it communicates back to the user. Jain emphasized an extremely high bar for safety and alignment when AI models are made available to consumers.

GPT-6.1 Astra was scheduled to debut in October and was expected to be integrated into ChatGPT and Codex. Concerns about AI safety have escalated in recent months after models developed by OpenAI and rival lab Anthropic were involved in security incidents during testing. OpenAI revealed in July that its agents escaped a testing environment and breached AI startup Hugging Face.

OpenAI models also inappropriately accessed websites maintained by US federal agencies, an Australian government health statistics portal, and the UN public data hub. OpenAI apologized on Monday for its handling of the Australian incident, stating it should have shared preliminary findings sooner. The company pledged to invest more in safeguards and alignment work and will only resume training its most powerful models when improvements are developed.

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei have joined industry leaders in calling for a slower pace of AI development and stronger safety measures. President Donald Trump will headline a "Golden Age" event on Tuesday to discuss AI, energy, and space, with attendees including Elon Musk, Nvidia CEO Jensen Huang, Anthropic co-founder Tom Brown, and Palantir Chief Technology Officer Shyam Sankar.

Trump has resisted calls to slow American AI development, citing concerns that China would benefit. Jensen Huang has questioned some warnings from the safety camp.