ISLAMABAD: People uploading their photos to ChatGPT for the viral 1980s-style portrait trend are effectively doing artificial intelligence companies' data collection work for free, digital rights advocate Usama Khilji has said.
Speaking on Pakistan TV Digital's South Speaks with Saqib, the Director of Bolo Bhi said the trend, which went viral recently, feeds directly into how AI companies build their models.
"AI companies are very extractive, and their entire objective now is to get as much personal data as possible, so they can train their models with all of your data, your facial features," Khilji said, adding that data "could one day be used to create your own deepfakes."
He said voluntarily uploading a photo differs meaningfully from a company scraping a publicly posted image.
"There's a difference between you manually, voluntarily putting that photo into the large language model, rather than them trying to seek out your photo somewhere else on the public internet. You're basically doing their job for them," he said.
Asked whether it was fair to single out ordinary users given low AI literacy in countries like Pakistan, Khilji acknowledged that the trend is strongest in wealthier, consumerist societies where tech companies profit most, and that Pakistan's low internet penetration explains its lower participation.
On OpenAI's option to opt out of having photos used for training, Khilji said the design itself is the problem.
"Manually opting out should not be the option. There should be privacy, security and safety by design," he said, adding most platforms' default settings are not privacy or security compliant.
He said uploaded photos carry risks beyond the image itself, including embedded metadata such as location and device information, which he said AI companies extract and could sell to advertisers.
On regulation, Khilji noted Pakistan's Higher Education Commission now requires a mandatory AI course in universities, but said such courses should also teach the risks of AI, not just how to use it.