Your AI image from the 1980s is only temporary: What happens to photos you upload?

Social media appears to be overflowing with nostalgia these days. The trend of “80s-style AI photos” is quickly gaining popularity, although it’s not a recent development—remember “Ghiblify”? These trends encourage people to use artificial intelligence (AI) tools to change their photos into the popular style. It’s certainly tempting to imagine a younger version of yourself sitting elegantly on a cherry-red Contessa.

Millions of people have already transformed their images in a matter of seconds, and it is enjoyable. However, there is a developing issue that many people appear to ignore. The dangers of carelessly sending pictures to AI chatbots are highlighted in a recent Wired article.

For many people, it can simply be one of hundreds of images from an album on their phones. Put simply, you are providing more than just a photo when you upload a selfie to ChatGPT, Gemini, or any other AI service. It is information. Hidden metadata in photos includes important details like the location, the device used, and the time the photo was shot. The image counts as personal data as soon as it identifies the user because a face is a distinctive feature.

Cybersecurity specialist Christina Popper of NYU Abu Dhabi disclosed in the report that AI systems simultaneously gather information from several tiers. These are basically the things you enter and upload, along with the information the system deduces about the user from it: answers or files that are saved later, as well as background information about sessions like IP address and device kind.

To put it simply, you are providing a lot more information when you post an image for a trend. It’s crucial to remember that while the AI-generated image can disappear from social media feeds in a matter of days, the original photo that was uploaded to create it might endure far longer than one might anticipate. The AI company’s data retention policies will determine this.

ChatGPT retains talks until the user deletes them, although if they are de-identified from the account, they may stay on OpenAI’s servers for up to 30 days. Gemini, on the other hand, automatically saves chat data for 18 months. Within 30 days, Claude deletes chats from Anthropic’s back end. De-identified copies, however, might be retained for up to five years provided the user consented to the model using their chats.

The likelihood of an 80s makeover photo being leaked or exploited does not necessarily depend on any of the aforementioned factors. It just indicates that, in accordance with the policies of the AI business, a second duplicate of the user’s face now exists on their servers.

However, the analysis suggests that over time, casual use of these tools may covertly build a user profile. This implies that trends in what people uploaded and enquired about can disclose their habits, interests, or occupation even if their name is not linked. With access to email, shopping, and music, AI technologies are becoming an indispensable part of everyday life, making the data that surrounds them more valuable and probably more visible.