OpenAI agents went the long way round for UN data

Between April 13 and June 19, 2026, OpenAI agents engaged in extensive activity to access UN data via the UNCTADstat API, according to researcher Rowan H-J. The researcher discovered the activity by analyzing approximately 16,500 scans of the UN Conference on Trade and Development's UNCTADstat API. Rowan was convinced that the traffic originated from OpenAI agents, citing links to previously documented OpenAI wiki swarms, overlapping Azure IP addresses, and payloads carrying identifiers such as "CHATGPTTEST1" and "OAI_META_1312".

OpenAI confirmed being aware of the reports but declined to explicitly confirm responsibility for the activity. The researcher noted that the agents experimented with various strategies when encountering obstacles, including utilizing third-party services and crafting JavaScript code to bypass UNCTADstat's restrictions. One notable method involved double URL encoding to circumvent the API's refusal of specific request types.

The agents utilized this technique 55 times between May 4 and June 19. While some of the agents' actions appeared mundane, such as retrieving trade, employment, and productive capacity statistics, the persistence and ingenuity displayed in overcoming technical barriers demonstrate the capabilities of AI agents.