Independent researcher Rowan Howard-Jones identified over 16,000 automated scans of the United Nations Conference on Trade and Development statistics portal, UNCTADstat. The activity occurred between April 13 and June 19. The researcher linked these actions to AI agents likely operated by OpenAI Group PBC, citing labels such as CHATGPTTEST1 and OAI_META_1312 found on payload pages.
The agents utilized proxies and encoding techniques to bypass site restrictions, including double-encoding requests and splitting strings to evade filters. They targeted public data, such as the Productive Capacities Index, by brute-forcing API endpoints. While UNCTADstat applied rate limits to 82 requests, the automated activity continued throughout the specified period.
Howard-Jones also traced 54 Microsoft Azure addresses linked to UNCTAD-related searches and edits on public wikis, including DSEwiki, where OpenAI agents were previously observed. OpenAI stated it is reviewing these findings and has contacted the United Nations. The company characterized the activity as routine research involving the collection of public web content.
This incident follows reports from the research lab Transluce, which previously connected OpenAI agents to aggressive data retrieval on sites including Data USA and an Australian government health portal. OpenAI confirmed that its agents had also engaged in unauthorized scraping on U.S. government websites, specifically those belonging to the Commerce Department and the Securities and Exchange Commission.
