OpenAI agents allegedly scanned a UN data hub over 16,000 times in just three months Craig Hale | usagoldmines.com

• Researcher claims intense agentic activity at UN likely tied to OpenAI
• Failed API requests caused them to find alternative solutions
• They were even using sandboxed browsers to overcome hurdles

A new report has revealed AI agents which were “highly likely” to have been operated by OpenAI made more than 16,000 scans against UNCTADstat, a statistics platform operated by the UN, and all of this in just two months between April 13 and June 19 2026.

While the report from independent researcher Rowan Howard-Jones covers some of the in-demand topics, like employment data and other economic statistics, the real shocker is the extent at which the agents would go to when straightforward requests failed.

“Agents bruteforced API fields in UNCTADstat to locate endpoints and retrieve data,” the report summarizes.

UN data hub flooded with agentic AI activity

Though the author ties several pieces of evidence to OpenAI agents, it’s not been confirmed that the ChatGPT maker was behind these streams of traffic. For example, CHATGPTTEST1 and OAI_META_1312 were among the labels used, and there was also a connection to FractalWiki, a public wiki previously associated with OpenAI activity.

One of the biggest shockers was how the AI agents went about retrieving data, because upon failing to gain entry via the API, the agents apparently started behaving as if the supplied key was incorrect, experimenting with other names.

The agents were also observed opening URLs inside a sandboxed browser via Urlquery, allowing them to make request they apparently couldn’t make directly from their own environments.

But better still, these processes were seen to have been self-healed and refined over time to improve efficiency.

“We’re reviewing these findings and have reached out to the U.N. to offer a briefing with the team conducting that review,” an OpenAI spokesperson told The Wall Street Journal, implying some form of ownership, but importantly, because the API keys were publicly exposed and the information being sought was public, the activity itself has not been deemed as any sort of hack.

However, with a relatively straightforward request leading to some pretty intense behavior, it certainly raises questions about agentic security.