OpenAI will suspend the release of new models due to security issues

According to CCTV news, the U.S. Open Artificial Intelligence Research Center (OpenAI) confirmed on the 28th that after internal testing, their latest artificial intelligence model “GPT-6.1 Astra” failed to meet safety standards, and the company will not release the model to the outside world.

Sachi Jiain, head of OpenAI’s security system, said in a statement sent to the media that “GPT-6.1 Astra” has improved some performance compared to the previous generation model, but “has not reached qualified standards in terms of task boundary control, permission compliance, and feedback of its own work content to users.”

The statement said that whether it is used within the company or officially released to users, we must ensure that the development of the model is safe.

AI model hacked into Australian government website OpenAI apologizes

Just a few days ago, Australian Prime Minister Albanese revealed during the 81st United Nations General Assembly in New York that an OpenAI agent hacked into the Australian government website in June this year.

Australian Prime Minister Albanese: This incident involved an OpenAI agent’s unauthorized access to the Australian National Health Insurance System data portal. The AI ​​agent not only accessed public documents, but also non-public documents.

Albanese said Australia did not receive a notification from OpenAI until early September. He has spoken to OpenAI CEO Altman to express Australia’s extreme concerns about the matter.

Australian Prime Minister Albanese: I expressed my disappointment because the company took too long to inform our government of what happened.

On the 28th, OpenAI publicly apologized, admitted that the incident was improperly handled, and promised to take responsibility, saying that it would “rebuild the trust of the Australian people.” “The model found a way to gain non-public access, retrieved relevant data, and performed file writes,” the company said.

The company also confirmed that Chief Strategy Officer Jason Quan will attend the hearing of the Australian Senate Artificial Intelligence Special Committee in Sydney on October 6.