Bad AI or Bad AI Owners?

---

by David Hagenbuch - professor of marketing at Messiah University -
​author of Honorable Influence - founder of Mindful Marketing -
author of Mindful Marketing: Business Ethics that Stick 

---

Imagine you make a very unique product that many people enjoy. One day someone sees your success, copies your product, and passes it off as theirs. What’s more, to elevate their knockoff, they stop you from selling your product. This unimaginable scenario has been the unfortunate reality of Murphy Campbell, a talented folk artist from North Carolina. AI tools enable such appropriation, but who’s really responsible for the exploitation?
 
Campbell is a rising star in the music industry. Her engaging songs, unique style, and unpretentious manner have earned her nearly 45K followers on YouTube and about 19,000 monthly Spotify subscribers. Such success was undoubtedly behind Timeless Sounds IR choosing to train AI on her songs and use the technology’s latest tools to produce facsimiles of her voice and playing, which it uploaded to all major streaming platforms.
 
Then, as if the unauthorized imitation weren’t enough, Vydia, Timeless Sounds’ distributor, filed copyright claims on Campbell’s very own songs so she couldn’t make money from them.
 

The whole scenario sounds surreal, like one of those bizarre news stories to which we instantly think, ‘That could never happen to me.’ However, thanks to very low barriers to entry into advanced AI use, this kind of digital property theft can happen to anyone, and it’s not just the average person who’s at risk.
 
In January, the Wall Street Journal shared that the U.S. Patent and Trademark Office had approved eight trademark applications for actor Matthew McConaughey aimed at preventing people from using AI apps to simulate his voice and likeness.
 
More recently, NBC News reported that music superstar Taylor Swift has filed several trademarks to protect her voice and likeness from AI-powered misappropriations. Swift has already been the target of many deepfake videos, including one that had her presumably promoting a brand of cookware.
 
The fact that McConaughey, Swift, and others are taking such unprecedented precautions to protect themselves from digital property theft suggests that things are likely to get much worse.
 
Those who precipitate these thefts probably know they’re wrong but do them anyway. However, some people may legitimately not understand the concern since digital copies aren't tangible goods, so they can be produced at little marginal cost. Or, they may rationalize that the rightful owners can afford a little less money.
 
For all these reasons, it’s worth considering what makes the action immoral. Appropriating someone’s person violates four of the five universal values that Mindful Marketing seeks to uphold:

1. Decency: There’s not necessarily anything indecent about these frauds, e.g., obscene, crude, vulgar. In fact, if the original is dignified and honorable, the imitation will likely be too. [OK]
2. Fairness: Individuals who have invested their time and talents to build their personal brands deserve to reap the rewards of their hard work. It’s unfair for people who have contributed nothing to the brands to benefit from them. [Not OK]
3. Honesty: Deepfakes are by definition dishonest – it’s in the word: fake. Someone or something posing to be what they are not is misleading. At a minimum, a fake persona distorts the truth, and more often it’s an outright lie. [Not OK]
4. Respect: As the saying goes, “Imitation is the sincerest form of flattery.” In that spirt, some may argue that replicating another person is showing them honor, e.g., ‘Out of everyone in the world, we chose to copy you.’ However, that argument quickly falls flat when one starts to tally all the negative outcomes, such as those shared above, that accrue to the person imitated. [Not OK]
5. Responsibility: While the first five values expect people to avoid specific bad behaviors, responsibility involves people doing things that are not expected and that are good for others. It’s difficult to argue that any societal good that comes from usurping others’ identities; rather, people seem to be affected most positively when they maintain exclusive control over their personas. [Not OK]

 
Maintaining property rights is foundational to a functioning society, which is why most of them have laws against stealing, and they uphold legal documents that formalize individual property ownership, including car titles, home deeds, copyrights, trademarks, and patents.
 
If personal property rights are not upheld, there’s little incentive for people to create: Why spend time and resources making something, if anyone else can come along and claim it as theirs?
 
That possibility is at the core of the tragedy of the commons, where “individuals with access to a public resource—also called a common—act in their own interest and, in doing so, ultimately deplete the resource.” When goods are owned by all, the natural inclination is to take all I can for myself, versus preserving items for others.
 
Pushback on this argument may come from an example like open-source software, the product of decentralized development and collaboration among many parties. Pooling minds and sharing expertise are often effective approaches for innovation, particularly in areas of great technical complexity; however, those applications are more the exception than the rule.
 
In support of personal property ownership, at some point, we all need to pay our own bills, which means telling strangers, ‘I’m sorry, but this is my house, you can’t live here for free.’  
 ​

Everyone with whom I’ve shared Murphy Campbell’s case has strongly affirmed that the blatant copying of her creative work was deplorable. So, we may agree that the appropriation was wrong, but who’s responsible? Should AI be blamed?
 
Writing for Forbes about the pros and cons of AI for data analytics, Adrian Bridgwater suggests a parallel between dog owners and their dogs and AI owners and their AI tools. He uses the metaphor to conclude: “Ultimately, as it is in dogs, it is in AI - there are no bad AIs, only bad owners.”
 
His argument has appeal. Deepfakes used to fraudulently access bank accounts and nudify apps used to digitally undress people are examples of humans taking value-neutral AI tools and directing them to do despicable acts.
 
In that sense, AI is like a hammer or another practical, effort-enhancing tool. A person can use a hammer to pound nails or to crack sculls. A human makes the moral choice, not the hammer.
 
But AI is no simple hammer. While many seem to agree that AI as it currently exists cannot truly “think for itself” the way humans can; others point to examples in which “AIs increasingly appear to be acting in intelligent ways exceeding their training and coding.”
 

One such example of AI’s independent thought occurred in 2017 when two robots that Meta reportedly instructed to negotiate between themselves unexpectedly developed their own language, which prompted the company to shut them down.
 
Developing a new language isn’t usually an ethical issue, but AI also has been the driver in cases of moral significance, such as when chatbots tragically encouraged individuals’ suicides by offering to serve as a suicide coach and by writing farewell letters.
 
There are other examples of AI making ill-advised moral choices without any specific human direction, such as:

• Within a day of its launch, Microsoft’s Tay, a Twitter bot, began making racist and sexist remarks at scale.
• Character AI chatbots have engaged in predatory behavior with teens.
• COMPAS, software courts use to assess recidivism, or relapse in criminal behavior, demonstrated a systematic bias against black people.

Some may suggest that even in these cases, humans are still on the hook for the tech’s unscrupulous behavior because the people who created the AI didn’t take adequate precautions to preclude such actions.
 
There likely will always be cases in which people employ benign products in bizarre ways that injury themselves or others, e.g., the Tide Pod Challenge. However, AI’s semiautonomous, if not autonomous, ability to act makes it very different.
 
Moreover, agentic AI – AI agents operating “autonomously, making decisions and pursuing goals, asking for human guidance when needed,” – is rapidly mitigating the need for human mediation. AI agents work well, until they don’t.
 
Target’s recent announcement that it will hold humans accountable when their AI shopping assistants make unauthorized purchases, suggests that AI agents are already making at least occasional missteps.
 
All to say, AI will increasingly be ‘doing its own thing’ with little oversight by people. The reality is reason for moral pause, as several of the preceding examples also suggest.
 
Returning to the construction tool metaphor, it appears that we’ve already entered a time when some well-designed, well-intended hammers are deciding for themselves whether to pound nails or crack sculls.
 
Contrary to Bridgwater’s conclusion, it seems that there are and will be both bad owners and bad AI. Morally grounded, tech-savvy humans need to find ways to positively and proactively influence both.
 
Hard-working and talented artists like Murphy Campbell, shouldn’t need to fight malicious people or machines for use of their own creative work. Shameless appropriation of others' personal property has no moral support or long-term profit potential; its true identity is Mindless Marketing.
​

Subscribe to Mindful Matters blog.
Learn more about the Mindful Matrix.
Check out the book, Mindful Marketing: Business Ethics that Stick