Is “Ethical AI” the New “Clean Coal”?
At the Serious Play Conference last week, a founder demoed their game designed to help kids with ADHD. The prototype looked promising: thoughtful mechanics, clear therapeutic goals, and, critcally, an AI-powered guide character that adapted to each child’s needs and progress. This AI guide, they explained, was the centerpiece of the game’s design and it’s therapeutic value.
“Ethical AI!” the founder quickly added, almost like a magical incantation that could ward off criticism.
The speed of that qualifier told me everything about how the industry thinks about ethics right now. AI, pause, ethical AI. As if adding the word “ethical” could dispel whatever concerns the audience might have about artificial intelligence working with vulnerable children.
I followed up with them afterward. They explained that their model runs locally on the device, doesn’t communicate with larger language models like GPT, and never records or transmits user data. These are genuinely good practices, especially when working with children. But that reflexive “ethical!” declaration reminded me of something: a similar pattern that advocates of potentially harmful tech have used in the recent past.
“We’re going to use oil, nuclear, natural gas, and coal. …CLEAN coal!”
The Clean Coal Playbook
Energy companies spent billions marketing cleaner burning processes and carbon capture technology while maintaining the fundamental business model of extracting and burning fossil fuels. The improvements were real but incremental. The marketing suggested transformation while delivering optimization.
The clean coal campaign worked by focusing attention on superficial improvements while avoiding systemic change. Coal companies invested in scrubber technology, but they never questioned the fundamental premise of digging carbon out of the ground and burning it.
I’m seeing the same pattern in AI. Companies hire ethics officers while maintaining engagement-first business models. They publish principles about user wellbeing while optimizing for time-on-platform. They announce “responsible AI” initiatives while keeping algorithmic decision-making opaque.
The improvements aren’t fake. Bias mitigation techniques genuinely reduce discriminatory outcomes. Safety research actually prevents some harmful AI behaviors. But when companies feel compelled to announce their ethics in product demos, I start looking for what they’re trying to distract from.
When Ethics Becomes Branding
Having designed AI systems for clinical populations, I can tell you the difference between ethics theater and actual ethical design. Real ethical choices happen in design meetings, not press releases. They’re the hard decisions about what features to cut, what data not to collect, what engagement tactics to avoid even when they work.
I’ve worked on AI that helps train radiologists to catch cancer earlier and systems that bring grandparents and grandchildren together through playful mediation. These projects felt genuinely helpful, not because we announced their ethics, but because every design decision prioritized user outcomes over our own metrics.
The companies building genuinely ethical AI aren’t the ones shouting about it. They’re quietly making choices that cost them short-term engagement but create sustainable, trustworthy relationships with users.
Meanwhile, companies practicing ethics washing abandon user-centered design the moment growth slows or competitors gain market share. When business pressures mount, the ethics initiatives get quietly shelved while the marketing copy stays unchanged.
The Structural Problem
But individual company choices don’t address the systemic issues that make AI ethics so complicated in the first place.
AI companies need massive datasets, often obtained without clear consent. They require enormous computational resources with significant environmental costs. Their business models frequently depend on capturing and monetizing user attention. Training data includes copyrighted material used without permission. The energy consumption of model training contributes meaningfully to climate change.
Just like coal companies couldn’t solve climate change while still mining coal, AI companies may not be able to solve these issues while maintaining current approaches to data collection, model training, and user engagement.
The regulatory capture angle makes this even more concerning. Major AI companies are defining the ethics conversation around issues they can control (bias mitigation, safety research) while avoiding systemic problems they can’t or won’t address (energy consumption, intellectual property concerns, economic displacement). When OpenAI or Google calls for AI regulation, they’re often creating compliance costs that smaller competitors can’t afford.
Where Clean Coal Breaks Down
The analogy isn’t perfect, though. Unlike coal mining, AI technology isn’t inherently extractive. AI can genuinely augment human capabilities, solve complex problems, and reduce rather than increase harm.
The question is whether current industry practices allow for these positive outcomes, or whether they systematically prevent them. AI development could theoretically become more sustainable. Smaller, more efficient models could reduce energy requirements. Training on synthetic or properly licensed data could address intellectual property concerns. Business models based on user outcomes rather than engagement could eliminate manipulative design.
But these changes would require abandoning practices that currently drive most AI company valuations. Can an industry transform itself when transformation threatens its core economics?
Academic research on AI alignment, fairness, and interpretability is producing concrete improvements. The regulatory environment is pushing beyond voluntary principles toward enforceable standards. Some AI applications genuinely solve problems without creating new harms.
The technology itself has potential. The question is whether the industry developing it can resist the clean coal temptation of allowing incremental improvements to substitute for necessary transformation.
The Real Test
Here’s how to tell the difference between genuine ethical AI and ethics washing: look at what happens when business pressures mount.
Clean coal companies abandoned environmental initiatives the moment they became expensive. Companies practicing ethics washing do the same thing with user-centered design when growth slows.
Companies building genuinely ethical AI maintain those principles even when they’re costly. When companies detect their competitors gaining an advantage by sweeping ethical design under the rug, will they be able to resist doing the same? Can they actually prioritize user agency over engagement metrics, when their bottom line is at risk?
Even if they do, the most ethical individual companies can’t solve industry-wide problems. The AI ethics conversation needs to address both company-level design choices and systemic industry practices.
Beyond the Marketing
The stakes here extend beyond any individual company or product. AI systems increasingly shape how people access information, make decisions, and understand the world around them. Systems that manipulate user behavior, amplify misinformation, or systematically discriminate can undermine democratic institutions and social cohesion.
The clean coal analogy illuminates something important: when an industry faces criticism for fundamental problems, the first response is often to rebrand rather than restructure. “Clean coal” became a way to continue business as usual while appearing responsive to environmental concerns.
“Ethical AI” risks becoming the same thing: a way for companies to continue prioritizing growth and efficiency over sustainability and user agency while appearing responsive to valid concerns about AI’s impact.
Most AI ethics initiatives fall somewhere between genuine progress and pure marketing. Real improvements that don’t address systemic problems. Genuine principles that get compromised under business pressure. Technical solutions that solve some problems while creating others.
The industry still has a choice. It can follow clean coal’s path of allowing incremental improvements to substitute for necessary transformation. Or it can embrace the harder work of building business models, regulatory frameworks, and development practices that prioritize long-term benefit over short-term growth.
The companies making that choice quietly, through design decisions rather than press releases, will determine whether AI becomes a genuinely beneficial technology or just another extractive industry with better PR.
And next time you hear someone reflexively add “ethical” in front of “AI,” ask what they’re really trying to sell you.
Sam Liberty is a gamification expert, applied game designer, and consultant. His clients include The World Bank, Click Therapeutics, and DARPA. He teaches game design at Northeastern University. He is the former Lead Game Designer at Sidekick Health.
