Punishing the Pioneers, Rewarding the Slop: Why Gaming Reviews Are Getting AI Ethics Wrong
If you needed proof that the current debate around Artificial Intelligence has lost all nuance, look no further than the gaming press last month. In a bizarre twist of logic, we have seen ARC Raiders punished for building a sustainable, compensated AI model, while Call of Duty: Black Ops 7 skates by despite shovelling what looks like generic "AI slop" onto its player base.
Eurogamer’s decision to award ARC Raiders a punishing 2/5, explicitly citing its use of AI text-to-speech as a negative, while giving Black Ops 7 a passing 3/5 is not taking a moral stand. It is a fundamental misunderstanding of how automation economics works. By failing to distinguish between ethical scaling and lazy replacement, industry critics are inadvertently incentivising the very corporate greed they claim to oppose.
The Right Way: ARC Raiders and the Licensing Model
Let’s look at what Embark Studios actually did with ARC Raiders. They didn’t scrape the internet for stolen voice data. They didn’t clone the voices of dead actors without permission. They hired human voice artists, recorded them, and, crucially, paid them royalties for the use of their data to train a model.
This is the distinction that matters. In the corporate world, we call this a "Human-in-the-Loop" architecture. It isn’t about deleting the human; it’s about scaling their output. In a live-service game where dialogue needs to react dynamically to thousands of variables, you cannot physically drag a voice actor into a booth for every potential line.
Embark paid actors for the specific right to use their voice as a tool. This solves a massive technical hurdle, creating dynamic dialogue that reacts to gameplay, without stealing labour. The actors consented, they were paid, and the tech simply scaled their output. Reviewers, however, panicked at the label 'AI' rather than assessing the fairness of the deal.
The Wrong Way: Call of Duty and Automation Theatre
Contrast this with the "Studio Ghibli-style" calling cards found in Black Ops 7. This is what I call Automation Theatre. It’s the worst kind of corporate laziness: using generative AI not to solve a complex technical problem, but simply to avoid paying a concept artist for a few hours of work.
The resulting assets (allegedly AI-generated) lack soul, polish, and intent. They are filler content designed to pad out a Battle Pass. This is the "AI Slop" scenario everyone fears, a race to the bottom where quality is sacrificed for speed, and human creativity is sidelined for a prompt box.
Yet, despite widespread player backlash and a campaign that has been critically panned, Black Ops 7 secured a higher score than ARC Raiders. The message this sends to the industry is dangerous: "If you use AI transparently and pay your actors, we will punish you. If you quietly flood your game with cheap generative assets, we might complain, but you’ll pass."
We Need to Judge Implementation, Not Keywords
As someone who has spent a decade implementing automation in the real world, I can tell you that "AI" is not a monolith. It is a tool. You can use a hammer to build a house, or you can use it to smash a window.
We are currently seeing a knee-jerk reaction where any use of AI is treated as inherently evil. This is economic illiteracy. We cannot stop the tide of automation, but we can shape the economic model that governs it. We should be championing models like ARC Raiders, where humans are paid for the training data, as the gold standard for the future of work.
If we punish the ethical pioneers, the only companies left will be the ones who don't care about ethics at all. We need to stop reviewing the technology and start reviewing the implementation. ARC Raiders built a tool; Call of Duty arguably built a cheat. It’s time we learned the difference.
We'd like your views:
- Does penalising studios that pay royalties for AI training data actually hurt the artists you're trying to protect?
- Why is "visible" AI innovation (like dynamic voices) judged more harshly than "invisible" AI cost-cutting (like generated texture assets)?
- Should reviews separate "Technical Implementation" scores from "Ethical Stance" opinions?
- Is the definition of "Lazy AI" simply anything that replaces a human task, or is it specifically poor quality output?
- How should the industry standardise "Fair Pay" for AI training data to prevent exploitation?



