OpenAI agents: suspicious scans on UN website

A security researcher found that OpenAI's software robots were a bit too nosy. Between April and June, these "agents" scanned the statistics site of the UN Conference on Trade and Development (UNCTAD) over 16,000 times. This was a "bruteforce" attempt, meaning they were trying to guess passwords or find vulnerabilities.

Rowan Howard-Jones, a security researcher, noticed this unusual activity. He then linked these repeated accesses to OpenAI's systems. We're not talking about a sophisticated attack, but more of a persistent "probing." It's a bit like a kid repeatedly ringing a mansion's doorbell, just to see who answers.

Thankfully, no, not quite. This incident isn't as serious as the hacks on Hugging Face or recent attacks on US government sites, which had far graver consequences. There was no data breach or service disruption here, so we can sleep soundly.

Still, it makes you think. AI agents left free to roam the web can stumble into awkward situations. Who's supervising these digital "explorers"? The UNCTAD site incident, scanned 16,000 times by OpenAI's agents between April and June, shows how easy it is to lose control. It's not a hack, but a wake-up call for the future of autonomous AI.