
A new McAfee study has found that AI models can identify where a photo was taken even when it lacks GPS data, location tags or captions. The findings raise privacy concerns, especially for users sharing vacation photos on social media platforms such as X, Facebook and Instagram. Scammers could misuse the technology to create convincing fraud messages.

Researchers tested 21,236 travel images on two open-weight AI models: Google’s Gemma 3 27B and Alibaba’s Qwen3 VL 30B. Neither model was given any location information. Gemma correctly identified the city and country in 87 per cent of cases, while Qwen achieved 91 per cent accuracy. The models analysed visual clues such as buildings, signs, shop fronts, architecture, lighting and food stalls.
Images of famous tourist spots were easier for the AI to place, while generic hotel rooms, beaches and rural roads were harder. The study warns that these open-weight models can be used without a subscription, making them accessible to fraudsters who could harvest holiday photos from social media and use AI to estimate locations for targeted scams.
Sources (2): timesnownews.com, ndtvprofit.com
This story was synthesised by AI from the 2 sources linked above. Methodology and corrections.
Updated: this story now draws on 2 sources.