With the future uncertain, how do legal and ethical concerns size up against the efficiency and ease of AI image generation?
As investigations mount over the creation and distribution of sexualized, AI-generated images, users and nonusers alike are raising new legal and ethical questions about our increasingly AI-generated future.
Concerns about the safety of AI image generation have grown worldwide after Elon Musk and xAI’s social media platform X was flooded with highly sexualized, nonconsensual photos of real individuals generated by xAI’s chatbot, Grok.
The United Kingdom and California responded by announcing separate investigations into xAI, while Indonesia and Malaysia outright banned the Grok app from their servers. With pressure mounting, xAI said in a post on X January 14, that they had implemented safeguards to “prevent the Grok account from allowing the editing of images of real people in revealing clothing,” in jurisdictions where it’s illegal.
The latest technological measures announced by xAI give hope that AI can become regulated. However, as users continue to find new ways to abuse the abilities of AI chatbots, legislation must evolve with it. Michael Grynberg, a law professor at DePaul University, said holding AI companies accountable isn’t always straightforward.
“It depends on the kind of claim that is being made,” Grynberg said. “There’s a lot of ground that just needs to be explored.”
When it comes to the creation and publishing of illegal content using large language models (LLM), which are the systems behind generative AI platforms, there is limited precedent to turn to. Section 230 of the 1996 Communications Decency Act protects online platforms, such as Facebook or Instagram, from being held liable for illegal content posted to their platforms. However, Grynberg pointed out two wrinkles in this legislation that could allow for the prosecution of AI companies.
“If an LLM is creating the imagery, then Section 230 maybe shouldn’t apply.”
Section 230 labels social media companies as distributors, not publishers, and so when something illegal is posted on their site, the user would be found responsible because they created and published the image.
When an LLM is prompted to create illegal imagery, Grynberg explained, and does, the company can now be viewed as a publisher and Section 230 no longer applies.
Congress passed the Take It Down Act in May of last year, which makes it illegal to post nonconsensual, intimate images as well as deepfakes, which are digitally altered photos or videos of someone, typically used to make the subject do or say something. It also forces companies to take down validated, nonconsensual imagery within 48 hours of a request, or face penalties from the Federal Trade Commission.
This act creates a new avenue, Grynberg added, for people to tell websites to take down unauthorized, intimate material about them, which extends to AI-generated images. While criminal penalties for those posting this content went into effect upon signing, companies will have until May to implement notice-and-removal systems to weed out illegal content.
While there has been heavy focus on the dangers of nonconsensual, sexualized and AI-generated imagery of late — concerns about AI image generation extend beyond this. Jessica Meharry, associate professor of design at DePaul, is concerned that the development of LLMs threatens the property and livelihoods of artists.
“So much of it has been built without any credit or compensation for all that work it has scraped off the artists,” said Meharry. “I think that’s deeply unethical.”
When AI models are trained, Grynberg said, they take in large amounts of data, some of which have gotten the creators into legal trouble.
“A number of copyright holders have sued the makers of AI models,” Grynberg said, “for the fact that in the training of their models, copyrighted material is used without permission.”
“There are certain kinds of conduct that even though you are making a reproduction of the copyrighted work, it’s nonetheless okay.” Content is allowed what is called fair use of copyrighted material, Grynberg said, “because we think that this is the kind of activity that as a society we think is beneficial.”
How the use of copyrighted material is deemed beneficial is judged along four factors: the purpose and character of your use, the nature of the copyrighted content itself, the substantiality and portion taken, and the effect upon the market of the original content.
In online spaces, search engines like Google have been able to navigate copyright infringement by arguing their use of copyrighted websites and content is transformative. This creates the search engine itself, and does not impact the market of the original content. The makers of LLMs, Grynberg said, are using the same reasoning in recent cases. However, he added, things are different.
“The argument is that unlike a search engine,” Grynberg said, “the LLM is a competitor for the original works.”
While the creation of a generative AI platform is transformative, the content it produces can become direct competition with the artists and designers it gathered data from.
“It waters down the creative process, and waters down the output,” Meharry said, “and then keeps people who are really good at that work from being able to make a livelihood.”
With the possibility of having your work stolen or outcompeted by the quickness and ease of AI image generation, Meharry and others are concerned for the future of the creative world.
“Some of the jobs that exist now won’t,” Meharry said, “but thinking through how we pivot and respond and take more ownership rather than just letting it be done to us is key.”
At DePaul, Meharry teaches her students about the importance of metacognition when dealing with AI.
“Thinking about your own thinking, how using AI affects the process,” Meharry said, “and then really being able to discern what it’s doing and whether that’s of value.”
In the classroom, her students explore how technology has constantly changed the framework of their professions, as well as how they can adapt and use advancements like generative AI to their advantage.
“They are very conscious of it, they want to learn and they want to make,” Meharry said. “They’re not interested in just sort of plugging in a question and having it make an image, that doesn’t satisfy that creative impulse.”
Loyola University student and artist Miles Nelson agreed there are ways AI can benefit artists and creatives in their fields.
“I think AI is gonna be super helpful teaching people,” said Nelson, “at least the basics.”
Through his work as a music producer, videographer and clothing designer, Nelson said he utilized AI to help him learn tips and tricks for new technology he acquires, as well as an organizational tool and an idea builder. While AI should be used as a tool to refine your craft, Nelson believes it shouldn’t be relied on to create art.
“If you’re relying on AI to do the work for you, I don’t think you can call yourself a true artist.”
Art is a skill and a dedication, agreed School of the Art Institute of Chicago (SAIC) student artist Alex Kane, that shouldn’t be replaced by AI.
“Art is something that takes time and skill,” Kane said, “…I have spent my entire life becoming an artist.”
When people can create pictures using AI image generators in seconds, Kane said, not only does it create a multitude of legal and ethical problems, it takes away from the commitment and dedication artists have for their craft.
“One thing going to an art university has taught me is having pride in my own craft,” Kane said. “I can’t imagine an AI artist would have that same kind of pride.”
Header Image by Alex Carrasquillo
Be First to Comment