Thousands of AI users simply can’t resist the urge to flood chatbots with lewd or sexually explicit requests, especially for image editing. At least in the case of Microsoft’s Copilot AI, none of your prompts are actually private.

According to 404 Media, Microsoft has hired hundreds of human contractors to review users’ image-editing prompts and uploaded photos.

The aim is to improve the product, and the process may well be successful. However, many of these workers say they have to review quite a few truly bizarre and even perverted sexually explicit requests – and that results in trauma.

Images and prompts the human contractors have to review and can’t unsee include upskirt photos or women in sexual positions.

Other users have asked Copilot AI to make foot fetish images of children’s cartoon characters, shorten a real woman’s skirt, or, more generally, simply put people into sexual positions.

She-Hulk and Ariana Grande

Curiously, human contractors of this sort weren’t even hired to accept or deny similar requests. As per 404 Media’s report, they’re only asked to review whether the generated image successfully fulfilled the prompt – for instance, whether it made the woman’s AI-enlarged breasts large enough.

Are there any guidelines? Not really. The report quotes internal Microsoft documents where contractors are told to rely on their “intuition.”

Trust your intuition – when you glance at the 2 edited images side by side, which one immediately feels like the better edit? Your gut reaction as a human viewer matters,

A set of instructions reads.

“When in doubt, go with your first impression. Human intuition is good at catching subtle quality differences that are hard to articulate.”

At least some contractors aren’t happy at all. One person who works on the prompts told 404 Media: “Faces are always uncensored, and many of the prompts are sexual in nature and dubiously consensual.”

Another contractor said they had to review pro-anorexia content and were “recoiled.” One request asked to make an image of Ariana Grande with anorexia, and a lot of users want She-Hulk images.

“Who is writing these prompts, and who is deciding that basically generating porn is what Copilot is now focused on? It’s hard to take things seriously when my focus has to be on which model generated the appropriate bust size or which middle-aged woman was put in the appropriate sexually suggestive position,” one contractor said in the thread on the internal message board, also obtained by 404 media.

Not just you and the machine

AI companies are often found to employ poorly compensated humans to run and improve their much-hyped tech.

For example, 404 Media also recently reported that OpenAI was hiring hundreds of mere mortals to read users’ prompts to improve ChatGPT’s responses.

And last week, it emerged that Meta’s new “personal assistant” AI agent Muse was using good old-fashioned human beings in call centers to handle phone calls to external businesses – and, obviously, didn’t want the outside world to find out about it.

Privacy is an added concern. Thanks to inspired marketing, AI users are usually confident they’re indeed conversing with an AI tool – but as it turns out, again and again, they’re actually providing, at times, ultra-sensitive information to a person.

A Microsoft spokesperson simply told 404 Media in an email: “Microsoft uses customer data as described in our terms of use, including to improve our products and enforce our code of conduct.”