AI poses a serious danger to personal data - study
A study reported by Popular Science warns that modern neural networks can determine a person's location from ordinary photos posted on social media by analyzing background details such as architecture, signs, vegetation, and vehicles. Removing EXIF data is no longer sufficient. Tests showed that AI can identify a street in London and a country from a forest road. Experts advise blurring recognizable details, cropping photos, and using private accounts.
Popular Science reports that advances in artificial intelligence are increasing privacy risks when posting ordinary photos on social media. Modern neural networks can determine a person's location—home, work, or vacation spot—by analyzing barely noticeable details in the background of an image. Previously, removing EXIF data was enough for protection, but now that is insufficient. AI analyzes architecture, signs, vegetation, terrain, cars, road signs, and taxi features. Tests showed that neural networks can identify a street in London based on buildings and vehicles, and even a country from a photo of an empty forest road. Such information could interest fraudsters and hackers for phishing, forging reservations, or tracking routes. Experts recommend hiding recognizable details: blurring signs, house numbers, and backgrounds, cropping photos more tightly for open profiles. They also advise limiting the audience, using private accounts, and sending personal photos through secure messengers. Earlier, Kursor wrote about the possibility of AI going out of control.