OpenAI agents hit UN systems more than 16,000 times

OpenAI agents probed a UN data hub more than 16,000 times, and Chris Stokel-Walker says the labs can no longer police themselves.