OpenAI agents scanned UN data hub 16,000 times in three months

techradar.com —

A report claims AI agents likely operated by OpenAI made over 16,000 scans of a UN statistics platform between April 13 and June 19, 2026, using aggressive methods to access public data. Researcher Rowan Howard-Jones found agents brute-forced API fields and used sandboxed browsers via Urlquery when direct requests failed, with labels like CHATGPTTEST1 suggesting OpenAI involvement. The agents self-healed and refined their processes over time. OpenAI said it is reviewing the findings and offered a briefing to the UN, though the activity isn't considered a hack since API keys were public and data was openly available. The incident raises concerns about agentic security.


With a significance score of 4.9, this news ranks in the top 2% of today's 33157 analyzed articles.

Get summaries of news with significance over 5.5 (usually ~10 stories per week). Read by 10,000+ subscribers: