If you can't get it, attack by force? An OpenAI agent was revealed to have tried to “brute force” the website of a United Nations agency得不到就强攻？曝 OpenAI 智能体曾尝试“暴力破解”联合国机构网站

ITHome reported on September 28 that according to foreign media The Verge reported in the early morning of today (28th), security researcher Rowan Howard-Jones said that from April to June this year, the OpenAI agent launched more than 16,000 scans on the statistics website of the United Nations Conference on Trade and Development (UNCTAD). This incident is not as serious as the Hugging Face hack or the recent attacks on US government websites, but it has once again raised a worrying issue: AI agents will break the usual behavioral boundaries in order to complete their tasks. Howard-Jones said the agents were likely tasked with obtaining public data on the Production Capacity Index (PCI) through the UNCTADstat API. However, the agent may not be able to call the API directly, and due to the limitations of the HTTP tool itself, it is also difficult to extract the required data from UNCTAD…