Moral philosophy is a big part of where my post-secondary education started, so this feels like a huge topic for me. I want to write forever, but I'll try to keep it constrained to what I see as the main issues: unethical training, eliminating jobs, undermining human agency, unsustainable business practices, and misleading or lying to the general public.
Unethical Training Big models take big data to train. The big models were all trained on as much data as they could get their hands on. This included quite a lot of copyrighted and use-limited material. Books, movies, academic articles, legal treatises, whatever it was the artists or authors were generally not consulted on whether their data could be consumed for this purpose. AI companies will argue they didn't break copyright, and many artists will disagree. But one way or the other they didn't have a chance to consent to this use. I'm not going to try to answer these because the conversation is much bigger than me and it doesn't directly threaten my livelihood. Instead I'll ask: IF mining those texts was immoral, what should we as individuals do about it? I'm focusing on individuals because it's in my zone of control and the law is not.
Should anyone be able to benefit (use the AI)? Should the companies who scraped the data be able to benefit (earn money from the AI)? Should the material producers be compensated?
If the material was obtained unethically, then the product is morally tainted. You can choose to use the result, not use the result, or try to replicate the product without the ethical problem. Even if you succeed at creating a "clean" version, you still have the problem that you were pointed at the solution by the unethical acts along the way. So there's no getting away from the moral residue. And since there's no getting away from moral residue, we're stuck where we are, with the harm already done, so our ethical duty becomes about what we can still do moving forward.
Well we could compensate the owners of the materials, but the harm was already done and we're talking about buying their consent, not their product. Since the harm was already done, the consent agreement is complicated because the ask is for retroactive consent. I'm not sure that can ever really be consent. Also I have no framework to compensate them individually. Maybe a really complex algorithm could be designed to figure out how much of a person's material I'm accessing and then I can pay for that, but from what I understand I don't think that's even possible. I'm sure there are mechanisms that the Intellectual Property world could apply. I suspect they'd have a lot of difficulty sorting out how much everyone contributed, but they could try.
We could demand redistribution of benefits, but that would be at the government level. I think this is actually pretty reasonable on a pay-it-forward basis. For instance if The Arts loses work, after providing a good deal of the content for AI to incubate, perhaps AI companies provide Arts funding. I can write a letter to my Member of Parliament, and in fact I think I will. The Arts may need help in particular because AI is now being used to try to do the job they have traditionally done. I don't think it will be successful, but it will nevertheless provide a deflationary effect to arts incomes, and many arts jobs will be lost.
I can individually limit my use of AI. That's certainly within my control. That's something people can and do choose to do, and it does slow AI adoption so there's a real benefit there. But it also introduces a tax on conscience. So even if most people agree that using AI is unethical you end up with a situation where the most moral agents are paying a price (lack of AI use and the efficiency it brings) for their morality. At that point, assuming AI is a force-multiplier tool, the most immoral agents have the most power. For instance you'll have bad actors gaining a disproportionate advantage over everyone else. This would break down very quickly so I don't think it's feasible.
And finally I can stop Google, Anthropic and Microsoft from benefiting from my personal use. This one actually is within my control. I can use decentralized/local AI tools with open models (not owned by those companies) and stay away from their pipelines. I'll pay an efficiency tax since they'll always be better at it than me, and I can also show other people how to do that, so that the denial of revenue to them grows. You can probably tell that last one is my favourite, but my reasons are more pragmatic than ethical. 1. I want to use AI. It's a powerful tool and I don't want to deny myself. That's not an ethical argument, it's a recognition of my bias. 2. This dovetails nicely with digital sovereignty both for Canada and for myself. 3. It's something I can sink my teeth into. I'm actively doing something. That's not morally better, it's just more psychologically satisfying. 4. me taking profits away from the big data scrapers doesn't fix the harm, but it reduces the benefit they accrue from my use. I feel that's worthwhile. 5. Perhaps a small moral improvement is that while doing this I can also choose to distance myself from those companies overall. So whatever Google was going to make from the fact that I used their browser or saw their ads is gone. Whatever they were going to benefit from me using a gmail account is gone now too. I suppose this is functionally a boycott in my home space.
Would it be better if AI didn't exist? Maybe. But that ship has sailed, so the remaining ethical choice is how I participate in the distribution of benefits. So I propose that decentralized AI provides a benefit over centralized AI models. And the best part of that is that we don't all need to agree. We can just do it individually. It doesn't require a critical mass to be meaningful at the individual level. I'm sure they won't miss me, but I can and am doing it right now. I'm not expecting others to follow this example, but I would like you to ask yourself: given where we are today, given the world you find yourself in, what ethical obligations remain to you? What can you do, or what must you do, to satisfy your ethical framework? I've found my answer. It's not clean, it's not settled, but it is mine.
Eliminating Jobs I think it's quite important that we separate the technology (AI) from the companies producing it (eg. Google, Microsoft and OpenAI), and from the users of the technology (us, our employers, our suppliers and customers).
There is a real risk here. I think work will change, but I don't think the ultimate result in this particular dimension will be as extreme as has been suggested. So far in 2026 we're hearing a lot about companies eliminating jobs to AI, but we don't seem to be seeing a lot of it. I'm sure that doesn't comfort anyone whose job was eliminated due to AI, but the scope currently seems to be quite small. That doesn't actually change the moral calculus though. If jobs are eliminated for this technology and new ones grow in their place, the work opportunities have changed, but that actually happens with any new technology. The net effect is critical but we don't actually know what that net effect is. The method of employment by corporations is very important, but it's not necessarily negative even though it seems inevitable that it must be so.
However, we don't look negatively at the word processor or the cotton gin. Word processors certainly eliminated typist positions because it was quicker to delete or alter text on a computer screen and keys didn't jam together. But I don't now imagine their net effect on work was negative. These tools are force-multipliers for human effort. I think there are very important elements to this, but they're not exactly where they may be expected. For instance, I think it's critical to maintain the responsibility for work with the company that employs the function. So if I get fired tomorrow and my company replaces me with MS Copilot they shouldn't be able to argue that any mistakes are copilot's mistakes. Offloading responsibility is actually where I think the job losses would come from. If you keep responsibility with a human, the human must be employed to hold that responsibility. This is basically a version or a consequence of "human in the loop" design. Maintaining responsibility with a human allows tasks to be shifted or altered while securing the worker's role. It doesn't, however, guarantee that the same number of humans will be employed or at the same kind of job. If only two of every ten employees (arbitrary figure) is needed to do the same work then over time the population will be de-mobilized, de-skilled and underemployed.
Quite an interesting effect that I'm not sure how to think about: introducing AI allows less skilled workers to do jobs that seem more-skilled. Overreliance would lead to a depressing effect on those same skills.
Undermining Human Agency
Removing responsibility from the human is several kinds of problem at once. Agency is the ability and also responsibility to make decisions and act on them. Another word for that is freedom. I'm not talking about AI slapping handcuffs on us, not physically at least. If you ask an AI "What should I do for my summer vacation" it does pattern matching and looks around for the things that are best represented in its data set or search parameters, and it recommends that. So when you ask a question like that you likely get a smaller distribution of activities than you would if some humans discussed it together. The information isn't wrong though. It's trying to answer the question you asked, but in doing so it's also encouraging you into a pattern of behavior that matches everyone else's pattern of behavior. The net effect, mathematically, would be more people doing the same few activities, leading to more people talking about those activities which makes the AI more certain that those activities are the right answer. We'd "suggest" ourselves into conformity.
Getting used to asking those questions also undermines the idea that you could or should make those decisions for yourself. Instead of asking yourself "what do I want to do", you take someone else's opinion. This isn't actually a new mechanism. This happens with newspaper's lists of activities happening this weekend in your area. It's how advertising or PR works as well. If I had to put my finger on why this is so objectionable to me, I'd say that it's because AI does it while pretending to be disinterested and representing fact. It's compounded by the degree of certainty you see in AI responses. We know that Advertising comes from the firm it represents. It's carefully curated words and images designed, usually, to sell us something. So in that sense it's honest. And to be frank we're used to dealing with those calls on our attention. We know that everyone wants to sell us things so we are prepared to resist those calls unless they prove themselves worthwhile.
At work, removing human agency means the humans may pay attention to only the problems the AI says are there. We may miss significant problems because they weren't told they're a problem. We're not using intuition guided by experience. I believe using AI doesn't automatically remove human agency, but it does introduce the risk. And the more we talk about AI having agency the greater the risk. An agent is one who decides and moves to act. An AI is never an agent. It's a software program with a bit of randomization thrown in. It makes a decision the same way that a flow-chart makes a decision. The flow chart sits there, pasted to the cork board until a human uses it. we don't talk about the flow chart deciding, we talk about "using the flow chart to decide".
The risk to human agency is significant, but I think it's also quite easily countered. Resist anthropomorphism and any language that suggests an AI makes a decision. Never forget that the humans do the thinking. Good or bad, right or wrong, we decide and we act. Our actions can be simple, or they can be complex. The knocking over of a single domino or a whole row of them. But either way, we act.
Unsustainable business practices
I think by now we've all read about how much power AI uses, and how new power plants are being built just to serve it. Sometimes, or even frequently, new uses of power are actually well worthwhile. They benefit humanity as a whole. Businesses are predicated on the idea that people want to buy whatever they're selling. But businesses have a responsibility to serve the public. The public does not have any such responsibility to support the business.
People need the energy AI is using. And I don't think there's any question that people must come first. The scale of AI operations means they concentrate a lot of capital in every data centre. They spend a lot of money, and that means taxes for their locality. So they seem to have influence over the political process through lobbying. If AI power use leads to an increase in power costs by citizens (and by the way an increase of carbon emissions) then AI is being subsidized by the citizenry. Regardless of its potential benefits, the AI industry must pay its own way. At the moment they are increasing their energy use massively. And even though they're huge and expensive, they're also very cost-sensitive: they use evaporative cooling because it's the cheapest cooling mechanism available at their scale. But again, people need that water. Farms need that water. If food or water becomes more expensive because of the AI industry then the citizenry is subsidizing the AI industry.
It's also not easy to justify allocating so much potable water to artificial intelligence when real human communities are suffering from lack of potable water.
Lying to or Misleading the General Public
Frankly this is the one that bugs me the most. That may seem bonkers, and if so I totally get it. If people don't get heat in the winter or safe drinking water, human lives are at stake. Nobody dies from being misled about technology. We could argue about whether AI companies are doing it deliberately. I'd suggest it's the result of a chain of decisions, where the technicians know how things work and they make statements intended to be comprehensible to their managers (losing technical accuracy with a little bit of magic), their bosses brief the marketing people (stripping out a bit more accuracy for magic), and the marketing people do it one more time. It all helps with getting across a single, simple message, but that message is now totally divorced from how the technology works and what it does.
Whatever the reason, it seems to me that this effect causes most of these other issues. Eliminating jobs? Companies will do that when they're convinced that the AI can do something better and cheaper than humans. That comes, in part, from marketing efforts that tell those companies that an AI can be an "assistant" or a "researcher" or "coworker". Companies think that an AI is reliable so they attempt to offload work and risk onto them. If they are convinced that the AI "makes decisions" they may think they can blame the AI when something goes wrong. That's a fundamental error and can easily be traced back to this misleading characterization. Undermining Human Agency? Would you settle your holiday plans on a web search alone? Or would you then check and see whether it has the experience you want? Whether you can afford it, how long it will take to get there? But people will act based on AI information, I content this is because they get the impression that a thought is taking place, or a decision is being made. It's an artifact of language that we can't fully escape, but we can guard against it if we try. Unsustainable business practices take place because of greed, sure, but also because of the nature of the tech industry. the growth expectation is such that billions are being invested in infrastructure. That's because of demand forecasts, and the demand forecasts are generated in part by what the AI companies can convince the populace about AI. It can do X, it can do Y. It will save you $Z. If those things aren't true, or aren't completely true, then misleading the public has led to this demand curve. If you stop representing AI as human-like thought, how much demand disappears? If you stop talking about how it's already passed the singularity, or how clever it is escaping the test environment you put it in, how much of that demand evaporates? It wouldn't kill the industry, but it could make it develop at a more manageable pace.
Conclusion
Now let's take a step back. I've listed out a lot of ethical problems with AI because frankly there are many of them. But the answer can't be condemnation alone. The ship has sailed. You can't un-break a vase or a heart. The question can't be "what shouldn't I do". That way lies paralysis and submission. You write your destiny, and that's half the point of the ethical problems I've discussed here. The responsibility and the opportunity and the agency are human. So it's not "what shouldn't I do" it's "what will I do?"
If something here sparks a thought, reveals a flaw, raises a question, suggests a new direction, or inspires an idea for a future article, feel free to send me a note. If you have relevant experience to share, I'd be glad to hear from you as well.
Helpful feedback may be incorporated into future posts, with attribution where appropriate.