By Chuck Gallagher — Business Ethics Keynote Speaker and Trainer
TL;DR Chuck Gallagher, AI ethics speaker and author, argues that the real direction of AI is being set by money and competition far more than by any philosopher. Companies, he says, only show their true values once that pressure actually arrives.
Back in 2018, Google made a promise, and they wrote it down. Their AI principles named four things the company said it wouldn’t build. Weapons, for one. Surveillance that ran past accepted norms too. That promise didn’t come out of nowhere. Google’s own people had protested its work on a Pentagon program called Project Maven, and the company backed off. So the principles were Google answering its own people, plain about where the line sat.
Then along comes February 2025, and just like that, the line was gone. Google rewrote the principles and cut the whole section that spelled out what it wouldn’t pursue — weapons language, surveillance language, all of it. A blog post explained the move, co-signed by the head of DeepMind, citing global competition and a world grown more complicated. Read reasonably enough, I suppose. Didn’t change that the promise was gone.
Who Is Actually Steering Artificial Intelligence?
A recent Guardian profile of the DeepMind philosopher asked a fair question. Whose moral compass should guide this technology? Good question. Trouble is, I think it may already have been answered, and not by any philosopher. Look, I spend my life in rooms full of people dead certain their values steer the ship. Certain right up until money and pressure walk in. Then they learn what was steering all along. Google’s 2018 promise didn’t survive a competitive market. Nobody there woke up itching to build weapons. The incentives shifted, the values bent to fit. Happens all the time.
Now, think about the money. Big Tech poured more than $400 billion into AI infrastructure in 2025. Projections push that toward $700 billion in 2026. Folks, nobody sinks that kind of money into a science experiment. That’s business money. It has to earn its way back — and that quietly changes the questions people even think to ask. You quit wondering whether a thing ought to be built and start asking how fast you can ship it. Nobody announces it. It just happens.
Have We Already Decided Without Deciding?
One reader wrote to the Guardian and put it well. He reached back to a 2010 thought experiment off a forum called LessWrong. Roko’s Basilisk, they call it. The idea goes like this: some future super-intelligent machine comes around to punish everybody who saw it coming and didn’t help build it. So your fear of tomorrow’s AI is what builds it today. Strange little loop, isn’t it? Anyway, his point was that we’re looking the wrong way. We keep bracing for a machine that hasn’t shown up, when the real force is already here — plain old competition, rivalry, the chase for a return.
And you know what? I think he’s right. We’ve decided without deciding. Nobody ever called a vote on whether AI should serve extractive growth or human flourishing — the market settled it by default, one earnings call at a time. That’s how the big ones get decided, and it rarely looks like a debate. More like a thousand little rooms nobody’s watching, where the easy answer and the profitable answer just line up.
What Are We Not Asking?
Here’s what tends to get lost. AI keeps getting sold as the answer to scarcity. Short on food? Ask the machine. But one letter writer caught the trap: what if part of that shortage is the machine’s own thirst? Fair question, if you ask me. These systems eat staggering amounts of water and power. Researchers at the University of California, Riverside, put out a much-quoted figure — a short back-and-forth with a chatbot can burn close to a bottle of water, just in cooling. So the very tool we’re building to fix the problem turns around and feeds it.
That’s the cost nobody puts up on the slide. And the roads we don’t take? They may matter every bit as much as the ones we do — you’d never guess it from the sales pitch. AI could help us live within our means. Or it could help us strip-mine the place faster and call that progress. Intelligence won’t decide which. People will. Whether they mean to or not.
Where Does This Leave the Rest of Us?
Let me be plain, because I learned it the hard way. A value you never get tested on isn’t really a value. It’s just a decoration. Trust me, I know what it is to tell yourself a story about the man you are, until a choice comes along and shows you the truth. Companies are no different. Google’s principles read just fine for seven years. Then a hard choice came, and the principles moved.
Look, I make a living speaking and writing about AI and AI ethics. Don’t get me wrong — I’m not saying distrust everyone creating this technology. Heck, I like this technology. But philosophy isn’t governance, so don’t make that mistake. Talking about where we’re headed is one thing, but making decisions under real pressure, well, that’s quite another. And that’s the decision that counts. So if intelligence as a commodity is about to get cheap and abundant, then the valuable resource is wisdom. The real value we’ll see moving forward is a human being’s critical thinking — and that’s worth guarding and developing.
Frequently Asked Questions
What did Google change about its AI principles in 2025?
Quite a lot, honestly. In February 2025 Google cut the part of its AI principles that named what it wouldn’t build — weapons and surveillance that violates accepted norms among them. Those lines had been there since 2018, put in after employees protested a military project. The company’s reason? Global competition and a messier geopolitical world. The commitments are gone now.
What is Roko’s Basilisk, and why does it matter here?
It’s a thought experiment from 2010, off a forum called LessWrong. Picture a future super-intelligent AI that punishes anyone who saw it coming and didn’t help build it. Most folks file it as a curiosity about machines that don’t exist yet. But one Guardian reader read it smarter: the pressure to race ahead on AI is already here, baked into present-day economics, long before any such machine shows up.
Are the people building AI ignoring ethics?
No. Not as a rule, anyway. I do this for a living, and most of the engineers and researchers I meet take these questions seriously — some lose real sleep over them. But here’s the catch. One person thinking hard on their own is a whole different animal from an institution making a call under pressure. When that pressure turns commercial, the fine principles have a way of bending.
Does AI really use that much water?
Oh, it uses plenty. More than most people would guess. Cooling the data centers that train and run these big models takes real water, and the electricity they pull carries a water cost of its own. The numbers bounce around depending on the facility and how you count. But the footprint’s big enough that a tool we’re selling as the cure for scarcity is quietly helping cause it.
So the lesson was never really about whether AI is good or bad. Direction is a choice, and it gets made by whoever’s paying attention when the pressure hits. As true in your shop as in any data center. Now, if your leadership is wrestling with where these tools should and shouldn’t take you — before the market decides for you — well, that’s a conversation worth having out loud. I’d be glad to bring it to your team, board, or conference. You’ll find me at ChuckGallagher.com.
Five Questions for Reflection
- Where has your organization written down a value it has never actually been tested on?
- When money or speed has collided with a stated principle in your company, which one won — and who noticed?
- Are you treating a thoughtful discussion about AI as if it were a decision about AI? What is the difference in your case?
- What opportunities might the current rush toward AI be causing you to overlook entirely?
- If wisdom is the scarce resource, who in your organization is responsible for supplying it?
Related Articles:
AI Ethics in Funeral Service: When Convenience Crosses the Line
