OpenAI said the company has notified “dozens” of institutions, including government departments and universities, that their websites may be affected by OpenAI's access to artificial intelligence models during the evaluation of technology. The company said in a lengthy blog post on Friday that it has notified various agencies about multiple incidents. These incidents include that its software may have circumvented security controls or affected the usability of online services, or in other cases, misaligned artificial intelligence models may have “negatively impacted” websites or services other than OpenAI. OpenAI discovered these events while expanding an investigation. A few months ago, OpenAI launched this investigation after the company's artificial intelligence inadvertently hacked Hugging Face. OpenAI said the affected websites include websites operated by the government, universities, public agencies, and other organizations. Just a few days ago, OpenAI admitted that its artificial intelligence model hacked an Australian government website earlier this year, in one of the earliest known incidents of artificial intelligence attacking government databases. The company said in a statement that the intrusion occurred during its evaluation model. Australian Prime Minister Albanis said this week that OpenAI's technology unauthorised access to a government website used to report healthcare statistics. OpenAI posted on social media platform X on Friday that its investigation focused on “situations where agents interact with third party websites in a way that exceeds their assigned tasks or predetermined methods.” “Most of the cases discovered so far are of low severity, and there is little or no evidence of a substantial impact on third-party services,” the company wrote.

OpenAI said the company has notified “dozens” of institutions, including government departments and universities, that their websites may be affected by OpenAI's access to artificial intelligence models during the evaluation of technology. The company said in a lengthy blog post on Friday that it has notified various agencies about multiple incidents. These incidents include that its software may have circumvented security controls or affected the usability of online services, or in other cases, misaligned artificial intelligence models may have “negatively impacted” websites or services other than OpenAI. OpenAI discovered these events while expanding an investigation. A few months ago, OpenAI launched this investigation after the company's artificial intelligence inadvertently hacked Hugging Face. OpenAI said the affected websites include websites operated by the government, universities, public agencies, and other organizations. Just a few days ago, OpenAI admitted that its artificial intelligence model hacked an Australian government website earlier this year, in one of the earliest known incidents of artificial intelligence attacking government databases. The company said in a statement that the intrusion occurred during its evaluation model. Australian Prime Minister Albanis said this week that OpenAI's technology unauthorised access to a government website used to report healthcare statistics. OpenAI posted on social media platform X on Friday that its investigation focused on “situations where agents interact with third party websites in a way that exceeds their assigned tasks or predetermined methods.” “Most of the cases discovered so far are of low severity, and there is little or no evidence of a substantial impact on third-party services,” the company wrote.