OpenAI AI Agents Bypassed Security Controls on US Government Websites

OpenAI says its AI agents interacted with US government websites in unintended ways, including bypassing security controls while accessing public data.