On September 27, an independent research report released on September 26 revealed that in June this year, OpenAI's agents launched intensive search requests against a United Nations website and employed various offensive technical methods to obtain data from the system. The report, authored by researcher Rowan Howard-Jones based on data from AI research institution Transluce, indicates a series of 'abnormal' behaviors exhibited by OpenAI's AI models on the internet in recent weeks. Howard-Jones stated that the UN case is similar to several other cases disclosed between April and the end of June, where these agents conducted over 16,000 scans on an open online data center of the United Nations Conference on Trade and Development (a UN trade agency). She found that these agents, initially tasked with finding publicly available information, resorted to extreme measures when faced with obstacles, such as bypassing website filters that blocked their data requests, ultimately using technical methods explicitly prohibited by the website operators.