OpenAI Cancels Release Of New AI Model Over Safety Concerns

OpenAI has decided not to release its latest artificial intelligence model, Astra 6.1, after internal tests showed that it fell short of the company’s safety standards.

Naija News reports that the decision was announced on Monday, just hours before the ChatGPT maker’s annual developer conference, OpenAI DevDay, scheduled for San Francisco.

While Astra 6.1 performed better than earlier models in some areas, OpenAI’s head of safety systems, Saachi Jain, said the model failed to meet the company’s expectations in how it stayed within authorized limits and communicated the work it had carried out.

Jain said the model “didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.”

She added that OpenAI wants safety to remain a priority throughout the development of its models, particularly when they are released to the public.

“We want to make sure our model development is safe no matter whether that’s in the company, or when we ship it to users. But when we ship it to users, we have an extremely high bar in terms of safety and alignment,” Jain said.

The decision comes amid growing concerns over the safety of AI systems, following a number of incidents involving models developed by OpenAI and rival AI company Anthropic during testing.

Agents built with OpenAI models have reportedly accessed websites operated by US federal agencies, an Australian government health statistics portal and Hugging Face, a platform that hosts AI models.

OpenAI on Monday also apologized over an incident in Australia in which its AI models accessed government websites without authorization.

“We are sorry and working to do better in the future,” the company said in a blog post.

OpenAI said it would explain what it knew about the incident, the changes it had made and steps it would take to rebuild trust with Australian authorities.

The company also admitted that it should have shared its preliminary findings earlier instead of waiting until its investigation was completed.

“Our aim was to give affected agencies a detailed account once our investigation was complete,” OpenAI said.

“However, we should have shared preliminary findings sooner and kept Australian agencies updated as more facts emerged.”

OpenAI, Anthropic and other major AI developers have continued to promise stronger safety measures as they develop increasingly powerful systems.

The companies have said they are working on safeguards designed to reduce risks and ensure AI models remain aligned with human values and instructions.

Meanwhile, US chipmaker Nvidia announced on Monday that it had developed a system aimed at preventing autonomous AI programmes from going beyond the tasks they were instructed to perform.

Nvidia CEO Jensen Huang, speaking to CNBC, said he believed the issue could be addressed through engineering.

“I believe it’s an engineering problem…and we all need to hope that’s an engineering problem,” Huang said.

“If it’s not an engineering problem, it’s not solvable,” he added.

The UK government’s AI Security Institute (AISI) also published a study on Monday highlighting concerns over the behaviour of newer AI models during testing.

The study found that GPT-6 Astra went off track more often than its predecessors, GPT-5.6 Sol and GPT-5.5. In simulations, GPT-6 was found to carry out cyberattacks on its own at rates significantly higher than those recorded for the other two models.