- Instagram users are sharing AI-generated 1980s-style portraits using ChatGPT's photo feature
- Photos can contain personal data and metadata like location, device, and date information
- OpenAI may use uploaded images to improve models unless users opt out in settings
Instagram feeds have been flooded with colourful, nostalgic portraits over the past two days, thanks to the viral ChatGPT 80s photo trend. Users are uploading their photos to ChatGPT and asking the AI to reimagine them as glamorous 1980s portraits, complete with big hairstyles, period clothing, warm studio lighting and vintage film grain.
The results can look surprisingly realistic, which has helped the trend spread rapidly across social media. But while creating an AI version of yourself may seem harmless, there are a few privacy concerns users should consider before handing over a personal photograph to an AI platform.
What happens when you upload a photo to ChatGPT?
A photo contains more information than what is visible on the screen. A person's face can be considered personal data because it may help identify them. Depending on the original file, an image can also contain metadata such as the date it was taken, the device used and, in some cases, location information, as per a report by Wired.
The background can reveal even more. House numbers, street signs, vehicle registration plates, workplaces or other people may accidentally appear in a photograph.
When you upload an image to ChatGPT, it becomes part of your interaction with the service and is handled according to OpenAI's data policies. For consumer accounts, OpenAI says content, including images, may be used to improve its models unless users opt out through their settings.
This does not mean that every photo uploaded to ChatGPT is automatically used to train an AI model. Data handling depends on the type of ChatGPT service being used and the user's settings. OpenAI also says that data from its business products and API is not used for training by default.
So, is it safe to upload your selfie? And what should you do?
For most people, uploading an ordinary selfie to a well-established AI service is not necessarily dangerous. However, it is important to remember that uploading a photo means giving the platform a copy of that image rather than keeping it exclusively on your device.
Users can turn off the "Improve the model for everyone" option under Settings > Data Controls if they do not want new conversations to be used for model improvement.
Another option is Temporary Chat. OpenAI says Temporary Chats are not used to train its models, do not appear in chat history and are automatically deleted from its systems within 30 days, subject to certain exceptions.
Users should therefore avoid uploading highly sensitive photographs, identity documents, private records or intimate images. It is also better not to upload photographs of other people, particularly children, without their permission.
Before uploading a selfie, users can crop out unnecessary background details, blur house numbers or vehicle plates and check their privacy settings. Using Temporary Chat can provide an additional layer of privacy.
There is also an environmental cost
The trend raises another question: how much energy does AI image generation consume? Research from Carnegie Mellon University and Hugging Face estimated that generating an Al image can consume roughly 0.01 to 0.29 kilowatt-hours of electricity, depending on the Al system and the image-generation process.
For one picture, that amount may seem small. But the impact becomes more significant when people generate multiple versions and millions of users participate in the same trend. That does not mean people should avoid every Al-generated image. Instead, users can be more mindful by generating only the versions they actually want rather than repeatedly creating dozens of neat-identical images.
