New AI scam is selling explicit images of someone who never existed
A new report has uncovered a new AI scam that swindles unsuspecting Reddit users into paying for explicit images of a woman who does not exist and is completely AI-generated.
The Washington Post has reported that a new AI-based scam has landed. Those responsible create explicit images using various, publicly known methods and post them on Reddit. The perpetrators then lead people to private messages, where they pay for AI-generated images of a woman who does not exist.
Despite the Washington Post approaching users on Reddit, only a few were perturbed by the news. One even said that the image still depicts a “pretty character”.
The report centers around “Claudia”, who we found through a quick Google search despite the Post’s censoring of private information. The account is quite active, and upon closer inspection of the images clearly indicates signs of AI generation.
Stable Diffusion part of explicit image AI scam
It appears that the AI image generator Stable Diffusion is the engine behind the scam. The image generator is easily downloadable and can be prompted to generate whatever the user wishes, regardless of the safeguards put in place.
Many of the images posted feature a white, simplistic room. This is easy for AI to create images with. However, most of the images have been posted in the 512×512 image aspect ratio, the default resolution for Stable Diffusion.
An image posted featuring blinds shows off one of Stable Diffusion’s biggest troubles right now. While it can generate blinds, when combined with the generation of the figure of a woman, the blinds can become warped or crooked.
Another image that brought “Claudia” to the attention of members of Reddit and the report in question depicts just a face. However, looking at the room behind it, it was quickly pointed out that a ceiling beam stops being generated after it goes behind the subject’s head.
Further investigations to prove that this is someone posting AI content is found in the various explicit posts that “Claudia” sells. ZeroGPT, a detector for AI, couldn’t distinguish some of the posts as being written by an AI, but for the lengthier posts immediately found them to be 100% generated by a language bot like ChatGPT.
The AI race marches on
With calls to stop the development of AI, it appears that there’s no stopping this billion-dollar train as more companies rush to integrate the technology into their own applications.
ChatGPT’s owners OpenAI have partnered with Microsoft to bring the language model to their Office and Bing software, while Google’s Bard has soft-launched for those who have signed up.
In another report, it was found that ChatGPT’s latest version, GPT-4, is more likely to share misinformation during tests. Italy has also moved to ban ChatGPT until privacy issues can be resolved. Nvidia is also enjoying its silicon being used in many of these applications, having sold over 30,000 GPUs to OpenAI.
