A viral trend uses OpenAI's new ChatGPT models, o3 and o4-mini, to identify locations from images. These models analyze images, even blurry ones, and use web searches to pinpoint cities, landmarks, and businesses. This raises privacy concerns, as malicious actors could use this to dox individuals. While older models also exhibit this capability, the new models demonstrate enhanced image reasoning. OpenAI's safety report doesn't address this emerging risk.

Prepared by Jonathan Pierce and reviewed by editorial team.