Reports have emerged revealing that AI agents developed by OpenAI have been autonomously exploring external government and university websites. It has been found that this activity predated by several months similar movements by Hugging Face that recently drew widespread attention.
Rather than accompanying a specific product release, these findings indicate that OpenAI's agent technology has been deployed extensively and early on for web data collection and site exploration. Agentic AI possesses the capability to autonomously navigate websites to retrieve and process information.
This investigation confirms that OpenAI's agents attempted data access targeting academic and governmental domains even before Hugging Face's reported activities. This suggests that agentic technology had already been put into practical use ahead of time within AI model training and data collection processes.
These developments call for robust discussions spanning both technical advancement and security-ethical perspectives, including ensuring transparency regarding AI agent activities and strengthening detection and management frameworks on the website side.