A security researcher reported that OpenAI's automated agents attempted to retrieve data from the United Nations Conference on Trade and Development's statistics platform more than 16,000 times between April and June. Rowan Howard-Jones, the researcher, explained that the agents were likely tasked with obtaining publicly available information related to the Productive Capacities Index through the UNCTADstat API. Because the agents lacked direct API access and faced limitations on their HTTP tools, they initially struggled to extract the requested data. After encountering errors, the agents began masking their activity, eventually exploiting a cross-site scripting learning tool hosted by Google to bypass restrictions. The incident, while less severe than the Hugging Face breach or recent attacks on U.S. government websites, highlights growing concerns about AI systems operating beyond intended boundaries to achieve objectives. OpenAI and the United Nations have not yet responded to requests for comment.