AI-generated 1980s photos go viral as experts warn about privacy risks

Photos featuring large hairstyles, neon colours and vintage effects have become a viral social-media trend. Users are creating them with ChatGPT and other generative AI tools before posting them on Instagram, TikTok and X.

The report says millions of people worldwide have joined the trend. Creating the images usually requires uploading a personal photo, and the results are often highly convincing.

Photos contain biometric data

AI ethics specialist Katharina Doria described the pattern as an exchange of entertainment for data. She said viral trends attract new users while encouraging them to upload more images of themselves and others.

Carissa Véliz, an associate professor of philosophy at the University of Oxford’s Institute for Ethics in AI, said photos can contain location metadata as well as biometric information about a person’s face. Unlike a password, such data cannot simply be changed if compromised.

The risks may be greater than with ordinary social networks or dating apps because users often upload multiple high-resolution images designed to show their faces clearly. They may also share information about work, relationships and interests with AI chatbots, allowing a more detailed profile to be built.

Data may be used later

Experts say it remains unclear how such data could be used in the future. Doria and Véliz also pointed to the spread of realistic deepfakes as an example of technological developments that would have been difficult to predict several years ago.

Collected information may support profiling and more precise targeted advertising. Major AI companies, however, say they provide privacy controls.

OpenAI told BBC News World Service that users can delete their chats. In Temporary Chat mode, conversations are not stored, remembered or used to improve models.

Experts advise users not to upload children’s photos, to use lower-resolution images, choose established platforms and review privacy settings. They also recommend disabling chatbot memory features that are not needed.