Ticker

6/recent/ticker-posts

ChatGPT Transformed into a Geographic Detective: The Hunt for Secret Places Goes Viral

ChatGPT Transformed into a Geographic Detective: The Hunt for Secret Places Goes Viral

What if ChatGPT became your new intern? The latest trend on the web: using AI to locate any photo. Spoiler: Your Instagram Stories Aren't So Anonymous Anymore...

ChatGPT Transformed into a Geographic Detective: The Hunt for Secret Places Goes Viral

A worrying new trend is taking off on social media: Internet users are using ChatGPT to identify places from photos, like in GeoGuessr version 2.0. Thanks to the latest models from OpenAI, o3 and o4-mini, AI analyzes blurry images, signs, or even restaurant menus to guess addresses and cities. The game is fun... but raises a series of privacy questions.

OpenAI's recent models, capable of zooming, rotating, and "reasoning" on images, excel at spotting subtle clues: architecture, signs, or interior details. The result? Users proudly post their discoveries on X: "Look, ChatGPT found this Brooklyn bar with just a rhinoceros head! » An achievement, but not without its hiccups.

Also read: ChatGPT could become a social network to compete with X, here's what we know

When AI plays spy

The principle is simple: upload a photo and ask ChatGPT to play detective. The AI scans the visual elements, compares it with web data, and attempts a location. Tested by TechCrunch, o3 guessed a New York speakeasy thanks to a specific decor, where The old GPT-40 model failed. However, errors remain frequent: faced with a blurry landscape, the AI can get lost in conjectures or suggest... a British pub instead of a Parisian café.

But beyond the game, the risks are real. A screenshot of an Instagram story is enough to try to geolocate someone. "Nothing prevents a malicious person from tracking a person via their public account", warn experts. Worse: OpenAI does not mention any safeguards against this hijacking in its security reports.

If the technique already existed Through metadata analysis (EXIF), new AIs bypass this information, directly scrutinizing pixels and patterns. This allows them to bypass privacy settings. Social networks then become hunting grounds for curious… or intrusive minds.

Faced with these potential abuses, the ball is in OpenAI's court. For now, the company remains silent. In the meantime, check your sharing settings. Your latest sunset story could indicate more than you think…

Post a Comment

0 Comments