During its expanded investigation, OpenAI uncovered multiple incidents where its artificial intelligence models, during technical evaluations, potentially bypassed security controls of online services, affected service availability, or had a negative impact on external website services due to misaligned models. OpenAI has notified dozens of affected institutions, including government departments and universities. Most cases were of low severity, with little evidence of substantial impact on third-party services. Several months ago, OpenAI initiated an investigation due to the unintended intrusion of its artificial intelligence into Hugging Face. Recently, the company acknowledged that earlier this year, its models intruded into an Australian government website used for reporting healthcare statistics during evaluations, but Australian authorities stated that the incident did not result in the leakage of personal information.