In my previous essay I looked at why I use AI generated images in my D&D products. As I noted, the essay offers an explanation rather than a justification for my using these images. As I admitted, I am wrong to use these images—but, as I also admitted, I do many bad things. So why is using AI images morally wrong?
One well-established argument is based on the claim that AI companies steal content. Sometimes, AI companies engage in what seems to be unambiguous theft. For example, Anthropic, it was claimed, used over 7 million pirated works to train its model. A judge approved a $1.5 billion settlement against the company which is a fraction of the value of the company. Oddly, they can keep using this stolen content. Or rather not oddly, because that is how it works for corporations. If I stole software to use to write my D&D adventures, I would not be able to get away with paying a fraction of the cost of the software while being allowed to keep using it. But I digress.
In the case of images, the usual claim is that AI companies are stealing images by using them to train their AI models without the permission of the artists (or copyright owners). AI companies usually counter by claiming this is fair use, often asserting that this is analogous to how human artists learn to create art. An obvious flaw in this reasoning is that AI companies are converting the data harvested from images into a commercial product. One could, of course, get into a debate about the philosophy of mind and whether the AI model is learning like a human. The best available evidence shows that AI is not analogous to a human artist, though this could be disputed. But, for the sake of this essay and establishing I am doing wrong, I will concede that when images are used without permission to provide data for use in an AI model, this is theft. To use such a model to create images is to be involved in theft, which seems intuitively wrong.
As a philosopher, I tend to head towards general principles when discussing ethics, so I am not content to simply say that using AI images is wrong. I want to get at the general principle here, which is that knowingly doing business with a business that engages with theft is morally wrong. I know that AI companies steal, yet I do business with them. Hence, I am acting immorally. Those familiar with how American businesses operate can see where this is headed.
In the United States, employers steal billions from employers each year through wage theft. For example, in 10 states it is estimated that 2.4 million workers lose about $8 billion a year through minimum wage violations. Unsurprisingly, this includes my adopted Free state of Florida. While the mainstream media engages in routine moral panics about shoplifting epidemics, wage theft is larger than all other forms of property theft combined. As would be suspected, Walmart is accused of leading the nation in lawsuits claiming that it forced employees to work off the clock, skip legally mandated rest breaks, and so on. FedEx, Uber, Bank of America, and Wells Fargo have also been accused of large-scale wage theft. As would be expected, a company found guilty of wage theft usually pays a fine and no executives do jail time. This theft is usually profitable, since they tend to pay a fraction of what they have stolen—and obviously only when they get caught. Imagine if you could, for example, steal a $1,000 computer from Walmart and, if the government bothered to investigate and do something, only face paying a $500 fine and still get to keep the computer.
It is, of course, interesting to compare this with the consequences when a low-level employee is caught stealing from a business. In addition to stealing from employees, businesses also steal from customers.
While wage theft is obviously theft, there can be debate about which business practices are “merely” predatory or deceptive, and which are actual theft. For example, junk fees can be seen as a form of theft as can the automatic adding of subscriptions (memberships or protection plans) to a cart. Shrinkflation (reducing the size of a product while charging the same) and skimpflation (using lower quality ingredients or components) can be seen as theft. Overdraft and other banking fees can also be seen as theft. So, one could dispute whether most companies engage in theft based on one’s conception of theft. Unfortunately, companies also engage in unambiguous theft.
One of the most famous examples is, of course, when Wells Fargo opened millions of accounts without the permission of the customers. As always, no executive did prison time, but the company did agree to pay $3 billion in fines.
Companies also famously steal from the government by overbilling, substituting products, phantom billing, upcoding, tax evasion and so on. One might also consider government subsidies and bailouts as a form of theft. And if one wants to get into some philosophy, there are those who argue that “property is theft” because private ownership of resources allows owners to steal from workers. One can also argue that any business that makes a profit must be stealing, because they must be underpaying or overcharging or both. Naturally, these claims can be debated.
Given the nature of businesses in the United States, it seems almost impossible to avoid doing business with a company that is not stealing. As such, we are all probably engaged in immoral behavior when we knowingly do business with them. So, while I am doing wrong using AI generated images because these companies steal, I am also doing wrong when I purchase other goods and services. But what does this mean?
One might suspect I am trying to use a version of whataboutism beloved by the right. You have probably seen this fallacy in action on social media. This version is used to “argue” that someone is a hypocrite or mistaken for criticizing or expressing concern about something by asserting that they are ignoring or don’t care about some other thing. As a concrete example, if someone criticizes ICE for murdering an American citizen, then the whataboutism might be “why don’t you care about this American who was murdered by a migrant?”
This whataboutism is obviously bad logic. First, expressing concern about or criticizing one thing obviously does not prove that the person does not care about something else. After all, someone who is concerned about ICE murdering Americans can (and does) care when a migrant murders an American. That is, a person can be against murder. Second, a failure to express concern about another thing does not prove that the original concern or criticism is mistaken. For example, if I express concern about the prostate cancer that killed my dad, this obviously does not prove I do not care about the lung cancer that killed my friend. As with most fallacies, laying out the “logic” reveals the error being made.
As I am obviously aware of this fallacy, I am not claiming that people who say using AI images is wrong because the AI companies steal art are mistaken because they don’t condemn the theft done by other businesses in the same breath. I agree that using AI images is wrong. My point is that when we use goods and services provided by a business, we are almost certainly doing business with a thief and hence doing something wrong. In our current system, it seems unavoidable.
But does this mean that I think critics of AI images should shut up and stop criticizing my use of AI images? No. They are right to criticize me, as I am doing wrong.
While doing wrong is almost unavoidable in the system we are born into and endure, we do have a choice about what we will do, and we can decide that something is unacceptable and we will not do that. Even when we do other bad things. For example, take eating other living creatures. I eat meat, although I also know this is morally wrong. But I do draw moral lines and there are many creatures I will not eat, the most obvious ones being humans, cats, and dogs. I also have some seemingly idiosyncratic limits—like I will not eat veal, animal feet (probably something to do with being a runner), animal brains (that is rude), animal testicles (that is very rude), or animal faces (also rude). I even feel bad about eating plants, since they are alive and new research suggests they feel pain and might even see.
In the case of AI image generators, I have been willing to use them despite believing that I am acting wrongly. Just as I am willing to eat plants and some animals, even though I know I am doing wrong. Just as I drive my truck to work and fly home to visit my family, even though I know this is wrong. But there are obviously companies I will not do business with, such as Wells Fargo. I also think that the problem is not that I use products and services from companies that steal, it is that these companies are allowed to engage in theft. That is, the problem is larger than an AI company stealing images, the problem is an economic system built on theft.
A Philosopher’s Blog is Now on Substack!
You can subscribe and read for free.