If AI is unethical, so is your phone
I’ve done a few webinars lately that have touched on using AI in the context of managing devices. Every time, without fail, someone shows up in the chat to mention AI is unethical. Sometimes it’s about the training data. Sometimes it’s about the environmental cost - the water, the electricity, the servers running somewhere I’ll never see.
I don’t think they’re wrong, exactly. I’ve argued the training data case myself. I’ve written about the actual number attached to the electricity and water. I hold these positions. I’m not writing this to defend AI from criticism.
I’m writing it to ask a different question: why is this the hill? Out of everything extractive and exploitative embedded in the tech industry, why is AI the one that gets you into the chat?
The mine you’re not thinking about
The phone in your pocket and the laptop you’re reading this on both run on lithium-ion batteries, and those batteries need cobalt. Over 70% of the world’s supply comes from the Democratic Republic of Congo. An estimated 40,000 children work those mines - some as young as six - often for less than two dollars a day, using their bare hands as tools because that’s the equipment they have. The scale of this is well documented. In 2024, a US federal appeals court sided with Apple, Google, Microsoft, Dell, and Tesla in a suit alleging they knowingly benefited from that labor. The ruling wasn’t that the labor didn’t happen. It was that buying cobalt through a global supply chain doesn’t make you legally responsible for what happens at the bottom of it.
Then there’s everything with a magnet in it - which is to say, nearly everything. Speakers, fan motors, hard drives, the haptic engine that buzzes when you get a notification. Rare earth elements, and roughly 90% of the world’s processing happens in China, a lot of it around a city called Baotou. Decades of unregulated extraction there left behind a tailings lake that’s been called one of the most toxic bodies of water on the planet, and documented contamination in the soil, water, and air of the communities around it.
Then there’s the lithium in that same battery - the thing that makes it rechargeable at all. Two mining companies pull more than 63 billion liters of brine a year from underneath Chile’s Salar de Atacama, and researchers have documented land subsidence and roughly a 30% drop in water levels in the salt flat, threatening water the Lickanantay people have depended on for centuries. This one has a particular irony to it: it’s happening under the banner of the green transition, the same “necessary for progress” framing that shows up every time someone defends building another data center.
Nobody brought any of that up in the chat. Not once. The person doing the complaining was typing it on a device built from the things they weren’t complaining about.
Three different countries, three different extraction methods, one common thread: the harm lands on people poor enough, and far away enough, that ignoring it costs us nothing.
What people are actually mad about
I have guesses about why the selectivity runs this direction - proximity, mostly, and the comfort of being able to point at something without indicting your own daily habits. But I didn’t want to just guess, so I went looking for actual numbers instead.
Gallup polled this in March 2026: seven in ten Americans oppose a data center being built in their area, nearly half strongly. That’s a real backlash - protests hit 125-plus locations across 40 states in a single day this July, and local opposition has already delayed or killed $130 billion in projects in the first quarter of 2026 alone.
So what are they actually mad about? Gallup asked, open-ended. Half mention resource use - water, energy, strain on the grid. About a fifth cite property values and quality of life. Another fifth cite their utility bill. Sixteen percent mention noise. Twelve percent are worried about their own job.
Ethics - privacy, misuse, safety - shows up too, but as a small slice of the total: four percent. I don’t think that means the ethical concern is fake, or that it’s secretly the environmental concern wearing a different word. They’re different objections, and both are legitimate on their own terms. I want to keep them as two separate threads instead of collapsing one into the other.
But pull on each thread separately and they lead to the same place. Look at what’s actually inside that four percent, and inside my webinar chat when someone says “unethical.” It’s training data. It’s an illustrator losing a commission. It’s a customer service worker losing a job to a chatbot. Real concerns, and I’ve argued for them myself. What’s never in there is the DRC. Baotou. The Atacama.
And notice what doesn’t happen next. Nobody who types “AI is unethical” into that chat follows it up by getting rid of the smartphone or the laptop they typed it on - the same devices built on the cobalt, the rare earth metalss, and the lithium. If that harm were actually what was driving the outrage, giving up the device would be the obvious next move. Almost nobody does it. Me included. The ethics gets aimed at the new thing walking in the door, not at the thing already sitting in their hand.
Even people raising a genuine ethical objection to AI are drawing the circle of who counts around people they can picture - an artist, a knowledge worker, someone whose job looks like theirs. It’s the same with the environmental concern: people worried about their own water table rarely extend that worry to a salt flat in Chile. Two different objections, and both of them stop at the same edge. That edge isn’t proximity exactly. It’s that we - and I mean this specifically, people like me, in wealthy countries - have built a habit of not caring about what happens to poor people in places we’ll never go.
The car nobody’s mad at
Someone posts about a new CarPlay feature and the comments are enthusiastic. Nobody’s asking about the emissions. Transportation is the single largest source of US carbon emissions - ahead of the power grid, ahead of industry - and the EPA estimates a typical passenger vehicle puts out about 4.6 metric tons of CO2 a year just from being driven to work and back. That’s not buried in a supply chain somewhere. It’s not abstract. It’s the thing almost everyone in that comment section does every weekday, and none of it gets brought up.
This might be the sharpest version of the environmental asymmetry, sharper than the data center one. A data center is new, visible, and can be pointed at from someone’s actual backyard. A car is old, familiar, and something almost everyone in the conversation personally owns and needs just to get to the job that pays for the phone they’re typing on. The emissions from a hundred million daily commutes are at least as real as the emissions from a data center - arguably a bigger, more durable problem - but they’ve had a century to become invisible. It’s hard to raise an environmental objection to something you did yourself this morning.
Not all AI is the same fight
I want to be careful here, because I don’t think “everything has a supply chain, so nothing is worse than anything else” is actually true, and it’s not the point I’m making.
Using AI to generate a customer service response that replaces a human worker is not the same as using AI to write a config profile. Using AI to approximate a working illustrator’s style without their consent is not the same as using AI to help debug infrastructure. I’ve written at length about where I draw that line and why - consent, and where the value actually flows. Those distinctions matter and I’m not walking them back.
But “not all AI use is equally bad” and “AI deserves uniquely intense scrutiny compared to literally everything else you own” are two different claims, and the second one doesn’t follow from the first. You can hold a nuanced position on AI specifically and still be applying a wildly inconsistent standard across your life generally. Both things are true of most of us, me included.
The selectivity is the point
There’s no ethical consumption under capitalism isn’t a cop-out, and it isn’t a reason to stop caring about anything. It’s a description of where we actually are. Nobody can fully audit the supply chain of everything they touch. You’d never leave the house. You’d never turn on a device. The DRC cobalt is in the phone of the person complaining about AI training data, and it’s in the laptop I used to write this blog post.
So we all draw a circle around who counts. Not out of malice, and not because the ethical concern or the environmental concern is fake - both are real, and I hold both. But the circle keeps landing in the same place: it includes people we can picture, people close enough to see, people whose loss registers as a loss because they look enough like us. It doesn’t include the 40,000 children in DRC cobalt mines, or the families next to the Baotou tailings lake, or the Lickanantay communities watching the Atacama sink. Those aren’t outside the circle because the harm is smaller. They’re outside it because they’re poor, and far away, and we’ll never have to look at them while it happens.
I don’t think the answer is to stop caring about your water table, or to stop raising real ethical objections to AI. I think the answer is to notice where your circle stops, on both counts, and ask honestly whether that edge was drawn by the size of the harm or by how easy the harm is to ignore.
Next time you’re about to go off in the comments, ask yourself who’s inside your circle and who isn’t - and whether that’s a line you actually drew, or one that was drawn for you by never having to meet the people on the other side of it. I’ll be asking myself the same thing.