The latest viral ChatGPT trend is doing ‘reverse location search’ from photos

There is some interest in connection with a new trend from viral: pictures use Chatgpt to understand the location shown in the pictures.

This week, Openai announced the latest AI models, O3 and O4-MiniBoth can “cause” unique through photos loaded. In practice, the models can make photos, rotate, rotate and enlarge – even blurred and distorted can be thoroughly analyzed.

Models paired with the ability to search for the Internet, make this image analysis capabilities, for a powerful place. Users in x are very good in the quick time O3, especially Cities, signsand even restaurants and restaurants from thin visual tips.

Wow, nail it and not a tree in the eyes. pic.twitter.com/bvcoe1fq0z

– Swax (@SWAX) April 17, 2025

In many cases, the “memories” of the former chatrpt conversations of the models do not appear or Exif dataMetadata attached to metadata to images that detect details as the picture is taken.

X is full of samples of users by giving CHATGPT Restaurant menus, Neighborhood Snaps, facadeand portraitAnd imagine “geoguessr” playing “Geoguessr” is an online game that has difficulty finding the images of Google Street View.

This is a fun chatgpt O3 feature. GeoGuessr! pic.twitter.com/hrcmixs8id

– Jason Barnes (@Vyrotek) April 17, 2025

This is an open potential confidentiality problem. There is nothing that prevents a bad actor of the screenshot, tell a person’s instagram story and using Chatgept to try them in Doxx.

o3 crazy
I wanted me to give me a random picture from a friend
They gave me a random photo they took in a library
O3 knows in 20 seconds and is correct pic.twitter.com/0k8dxifkoy

– Yumi (@ismuumi) April 17, 2025

Of course, this can be done before the O3 and O4-mini start. TechCrunch, a number of photos, GPT-4O, GPT-4O, GPT-4O, GPT-4O, GPT-4O fled a number of photos, GPT-4O to compare the techcrunch of models spatial guings. Surprisingly, GPT-4o received the same, correct answer O3 has received more frequent and less time.

O3 There was at least one example in our short trials when a place could not find GPT-4O. Taking into account a picture of a purple, built-in Rhino, Dimly-4o was not like a British pub, according to a Williamsburg Speakeasy.

From this point of view, you do not offer O3 perfectly. Several tests failed – O3 stuck in a loop, can’t come to a reasonable confident or wrong place voluntarily. In the X, users also celebrated that O3 could be pretty distant In place of space.

However, the trend shows some developing risks presented with talented, so-called-based AI models. ChatGpt has several security ways to prevent such as “reverse space search” and the company behind Chatgpt, does not apply to this issue Security report O3 and for O4-Mini.

We reached Openai for comment. If they are answered, we will update our piece.

10:19 PM Pacific: A few hours after this story was published, an Openai spokesman sent the following statement:

“Openai has designed to develop a visual thinking of O3 and O4-mini, developing instructions for the accessibility of emergency sensitivity, research or sensitive information, designed the instructions to determine the instructions and use of our rules to use our rules.”