In a recent unsettling turn of events, AI-generated explicit images of the renowned singer Taylor Swift have flooded X (formerly Twitter), illustrating the darker side of AI capabilities. These images, believed to be created using Microsoft Designer, garnered widespread attention and highlighted the ever-growing challenge of AI-generated fake pornography.

As these images rapidly spread across the platform, the incident not only sparked outrage among fans and privacy advocates but also raised critical questions about the ethical use of AI in media creation.

Michal Salát, Avast Threat Intelligence Director, brings a wealth of knowledge and expertise to this timely discussion. With a keen eye on the intersection of technology, security, and ethics, Salát offers a nuanced perspective on the implications of AI-generated content.

In the following sections, we delve into his insights, exploring the historical parallels of image manipulation, the ethical dilemmas presented by AI capabilities, and the broader implications for digital trust and security in our increasingly AI-driven world.

Fake porn is not a new issue

Salát draws a parallel between the current capabilities of AI in generating lifelike images and the use of Photoshop in the past. Just like Photoshop revolutionized image manipulation years ago, AI has now made it significantly easier to create realistic images. However, the concept of manipulating images isn't new; AI has just become the latest tool in this ongoing saga.

"This is just a different photo editing software, in a way," he explains, emphasizing that the concept of manipulating images has been a concern for over a decade. The key difference now is the ease and accessibility brought by AI technologies. This shift raises important questions about how society adapts to and regulates these new tools, which have made the creation of realistic images more accessible than ever before.

The ethical dilemma of AI and explicit content

One of the most pressing concerns is the ethical implications of using AI to generate explicit images, particularly of specific individuals without their consent. While generating generic nude images might not raise as many ethical questions, the real issue arises with the ease of creating explicit images of identifiable people.

“There’s probably nothing inherently wrong or ethically problematic with AI-generated porn,” Salat says. “But the ethical problem, at least for me, is that you can relatively easily generate an image with a known or a specific face on it."

This capability extends beyond celebrities like Taylor Swift to potentially anyone, emphasizing the need for ethical guidelines and legislation: While so-called “revenge porn” (or nonconsensual sharing of nudes) is illegal in 48 states, AI-generated deepfake pornography is currently only banned in 10 states. It's a stark difference that underlines the need for legislation to catch up with the technology.

How can AI companies prevent the creation of deepfake porn?

Perhaps because access to generative AI is a relatively new phenomenon, there are few guardrails around creating explicit content without the subject’s consent. According to Salát, however, implementing them wouldn’t be terribly difficult for AI companies to do: Something as simple as not allowing image generation of specific people or that uses customer-submitted images as a source material would provide sufficient barriers to the vast majority of people.

“Of course, some people will be able to break out of the restrictions and force the model to do something it’s not supposed to do,” Salát says. “The way I see it, at the moment, it’s virtually impossible to avoid this—you can only make it harder. I think the point I’m trying to make here is the companies that offer these services should try harder to avoid this.”

There’s also the possibility of someone creating their own porn-generating model that doesn’t have any restrictions. Salát points out, however, that even though it’s possible, it would take a lot of technical knowledge and a lot of computing power to do so.

“There aren’t a lot of people who would be able to create an AI to generate nudes themselves,” Salát says. “So, it would be more similar to using Photoshop, where you need to have a certain level of skill to be able to do that.”

Salát draws a parallel between the current state of AI development and the evolution of the security industry. He observes that we are witnessing the security industry "start over with AI." This comparison underlines the importance of viewing AI development as an ongoing process, where identifying and fixing vulnerabilities is key to maturing the technology.

The AI-generated images of Taylor Swift on X underscore the importance of responsible AI use. As AI continues to advance, balancing technological capabilities with ethical considerations and security measures will be crucial. By learning from experts like Michal Salát and reflecting on past challenges, we can navigate this complex landscape with a more informed and cautious approach.