1980s AI Photo Prompt ChatGPT: What Happens To Your Photos When You Upload Them To ChatGPT?

Add as a preferred source on Google

Open Instagram this week, and it is impossible to miss: big teased hair, mall-studio lighting, film grain, and a caption announcing someone's "senior year, 1986." The ChatGPT 1980s photo trend has taken over feeds across India, the US, and the Middle East, with Google Trends listing it among the week's top searches. The idea is simple: upload a current selfie, ask ChatGPT to reimagine it with retro hair, wardrobe, and studio backdrops, and share the result.

Photo Credit: Instagram: @parineetichopra @ravikishann

What most people do not pause to ask is where that selfie goes once it leaves their phone.

The Fun Part Takes Seconds. The Data Trail Takes Longer

Deleting specific conversations, memories, or an account is possible, and OpenAI outlines these options in its privacy policy. But by default, on free, Plus, and Pro personal accounts, uploaded photos are treated the same way as text prompts - stored, processed, and, unless the user actively opts out, used to help train future models, a setting managed under Data Controls in the account's Settings menu. Turning that off does not erase what has already happened: the opt-out only applies going forward, and content already folded into a completed training run cannot be pulled back out.

Even chats a user deletes are not gone immediately. OpenAI typically schedules deletions within about 30 days, though some deleted logs are being retained longer because of ongoing litigation.

Why A Face Is Not Just Another File

A selfie is not like a PDF or a spreadsheet. Tarun Pathak, Research Director at Counterpoint Research, has cautioned that viral AI portrait trends let participants unknowingly hand over rich digital profiles of themselves. He described the pattern as a steady feed that keeps benefiting AI systems long after a trend has faded from people's timelines, warning that the eventual risk includes highly targeted phishing built on stolen biometric details.

Shuya Feng, a cybersecurity researcher and assistant professor at the University of Alabama at Birmingham, has made a related point about what these models actually absorb. Beyond the image itself, she noted, a model studying a photo can pick up biometric particulars such as eye colour and hair colour; details with value that outlast any retro filter.

That distinction, biometric versus ordinary personal data, is what worries privacy researchers most. A password can be reset. A leaked fingerprint or a facial map cannot.

Small Steps That Can Help

Before uploading a photo for the next trend, a few checks make a real difference:

• Go to Settings > Data Controls and switch off "Improve the model for everyone" to stop new uploads from being used for training
• Use Temporary Chat for anything image-related - these conversations are not saved to history and are not used to train models
• Strip location and device metadata from a photo before uploading it, since that data can travel with the file
• Avoid uploading images that show number plates, home exteriors, ID cards, or children's faces alongside your own
• Treat any uploaded photo as something you can no longer fully control, regardless of what the settings say

Takeaway

The 1980s aesthetic will fade from feeds within a matter of weeks, the way every viral filter eventually does. The data behind it does not run on the same clock. A face, unlike a password, cannot be changed once it is out of your hands - which makes the upload button worth a second thought, long after the trend itself has moved on.