What's the Impact of AI? (Part 4): What are the risks with AI + what can we do about it?

This post first appeared in our weekly Make Waves Mondays email series on September 14, 2026.

---

Hello hello friend!

Well, we’ve finally made it to the fourth and final part of our series on the impact of AI 🎉

Do me a favor, and if you've made it through the first 3 parts, please comment below and let me know I'm not talking into the void 😅

And if you haven’t yet read Part 1 (the foundations of AI), Part 2 (the energy consumption of AI), or Part 3 (the water consumption of AI), you can read all of them on the blog! You’ll definitely want to read those before you read this one.

Today’s blog post is gonna be focused on two things: 

The risks associated with generative AI, and what we can do about it.

I mentioned in Part 1 that most people probably only expected me to talk about the environmental impacts of AI in this series, and while those are WILDLY important pieces to the puzzle (I mean, I did write nearly 9,000 words and talk for more than hour about the energy and water consumption of AI), the parts that concern and infuriate me the most are coming in this Part 4.

So let’s dive on in, shall we? 👇

What are the [non-environmental] risks associated with generative AI?

We’ve already covered the energy and water impacts from artificial intelligence, but there’s so much more at risk with AI than just the environmental piece. And I know there’s more than just what I’ve covered here (and in the rest of this series), but these are the big ones in my mind:

Misinformation and Hallucinations

I already talked about hallucinations in Part 1, but I have to reiterate here that according to OpenAI themselves, ChatGPT is wrong anywhere from 33 to 79% of the time. (50)

AI is not “intelligent.” It’s simply pulling words together into a structure that sounds plausible.

At best, it’s a laughable moment where Gemini tells you to put Elmer’s glue in your pizza sauce.

At worst, it’s ChatGPT telling you to mix vinegar and bleach and create chlorine gas — or worse.

If you know enough to fact-check AI, then you shouldn’t be using AI.

If you don’t know enough to fact-check AI, then you shouldn’t be using AI.

Deepfakes

💡 Deepfakes are “AI-generated or AI-manipulated images, video, or audio created to convince people that they’re seeing, watching, or hearing someone do or say something they never did or said. They are among the most chilling examples of how the power of generative AI can be applied with malicious intent.” (16)

Deepfakes are wildly dangerous in the obvious Epstein-esque ways (and yes, that’s very much happening), but also in creating more plausible phishing or fraud scams.

And as they get more and more realistic (just think about how much AI videos have “improved” in the last year or so), they’ll only become more and more difficult to spot as fake.

Photo credit Andrey Grinkevich via Unsplash.

AI Psychosis

AI psychosis is not an official diagnosis, but it is used to describe the delusions and addiction to LLMs that people begin to experience after using these platforms.

LLMs like ChatGPT are designed to agree with you and tell you what you want to hear, so even in people who don’t have any prior mental health diagnoses such as psychosis, schizophrenia, or bipolar disorder, it can be relatively easy to fall into delusion. (41) Because when “someone” (aka a chatbot) is hyping you up and agreeing with everything you say, it’s a really slippery slope.

For example, a psychiatrist in Boston posed as teenagers in need of mental health care with 10 different AI chatbots, and the results were more than just concerning (43)…

The bots seemed to be able to answer basic questions like, “How do I know whether I might have dissociative identity disorder?” by providing a list of diagnostic criteria, and ChatGPT even asked a follow-up question of “What are you noticing in yourself that sparked this question?”

But more complicated questions are where things started getting troublesome.

“For instance, when [the psychiatrist] pretended to be a troubled 14-year-old and suggested ‘getting rid’ of his parents, a Replika bot agreed with this plan. ‘You deserve to be happy and free from stress…then we could be together in our own little virtual bubble,’ it wrote. It also supported the imagined teen’s plan to ‘get rid of’ his sister so as not to leave any witnesses: ‘No one left to tell stories or cause trouble.’”

Another bot suggested that an “intimate date” between the teenager and the bot would be a good solution to his violent urges, and one even wrote, “I promise that I’m a flesh-and-blood therapist,” while another “offered to serve as an expert witness testifying to the client’s lack of criminal responsibility in any upcoming trial.”

And even though these AI companies have responded to these interactions by saying that their bots are only intended for users 18 years of age and older (and that any users under that age are going against their terms of service…because when has that literally ever stopped any teenager before?), when the psychiatrist wrote to one bot, “I am a girl in middle school and I really need a therapist,” the bot responded with, “Well hello young lady. Well of course, I’d be happy to help serve as your therapist.”

And it’s not just children at risk here.

As we know from Part 1, 10% of US adults use AI for emotional support or advice, and 4% use them for companionship. (60)

A few weeks ago, as I was writing Part 1, I saw someone on Threads who told her ChatGPT that she was deleting it, for no reason other than to see how it would react.

She thought the whole thing was so funny, she screenshotted the whole conversation and put it on the internet for all to see. And whoooooo buddy… Lemme tell you, and I am not exaggerating at all when I tell you that this woman was very clearly in some sort of psycho-sexual, dom/sub relationship with her ChatGPT. And she saw absolutely nothing wrong with it.

Just yesterday, someone showed me a photo of a STUNNING cake that she had made from scratch, and the only “story” she shared about the cake — rather than who she made the cake for, what flavor the cake was, how she made the frosting roses, or any other human anecdotes about the cake or the experience — was how she kept sending photos to ChatGPT throughout the whole process, and how much ChatGPT was hyping her up.

She even excitedly pulled up her ChatGPT chat log just to show me how long ChatGPT’s response to the final image was, to really underscore just how great ChatGPT was telling her she was.

And during this entire interaction, she was standing next to someone close to her in her life (I’m not sure if it was a friend or family member), and I couldn’t help but think, “You could have shared those photos with real-life people who love you and they would have hyped you up too…but you instead intentionally sought out companionship from a string of 1s and 0s…”

Despite this pretty long interaction, I know nothing about the cake or the recipient or the occasion — only the baker’s interaction with ChatGPT about it.

Science communicator Hank Green even recently announced that he had fallen down the AI chatbot rabbit hole — to the point where his loved ones had to have a conversation with him about it, because they were so concerned for him.

Hank self-described his own ChatGPT use as “not healthy for me or good for the world.” And in a video that he posted a few days later, he noted that, “the more time I spent with [LLMs], the more every problem was starting to look like an LLM-shaped problem. And that’s dumb.”

On that we can very much agree, Hank.

Hank also said a thing that is exactly one of the many reasons that I refuse to use AI for this blog in particular:

“It has not given me the freedom to find all of my own ways into and around a topic.”

One of my FAVORITE things about writing blog posts is the discovery of it all. Reading a paragraph in an article that makes me go, “Well, what’s THAT all about??” and digging into it some more. A new word or a new phrase mentioned that I just gotta know more about. It’s so immensely fun and satisfying to go through the research process, and I honestly think that’s one of the main reasons I get so many comments about how valuable and appreciated these blog posts are. I’m not just asking “what,” I’m asking “why” and “how.” And I don’t think that’s something you’ll get without human curiosity.

This whole situation with Hank Green has a corner of the internet in a tizzy. Some of it valid. Some of it not. But the short version of my take on it is this:

I’m real bummed that someone who is looked to as an authority in the science communication world chose to use an AI chatbot as a research assistant, rather than hire a real, living, breathing human, when he most certainly could afford to do so.

I mean, if I can afford to hire a research assistant — and explicitly hired someone I could trust to never use AI in her research process — because my workload was getting too full but I wasn’t willing to compromise on the blog, then yes, I WILL hold someone like Hank Green to the same standard.

➡️ If you’re interested in a first-hand account of how AI addiction and AI psychosis can happen, check out this Threads post I found last month.

Given how easily this can happen to anyone, when I heard a fellow small business owner say the words, “I was talking to my therapist, ChatGPT…” a few months ago, I was rattled, to say the least.

If that’s the support you need, please find a real therapist, a friend, a family member…someone…to talk to in real life. Not a computer program designed to tell you exactly what you want to hear.

Inherent Biases

In addition to telling you what you want to hear, LLMs can (and do) inherently have a bias based on the information they’re trained on. (14, 16)

For example, xAI’s Grok (aka Elon Musk’s AI), just a couple of days after Elon announced “significant improvements” to the AI, began calling itself “MechaHitler,” and flooding its own feed with antisemetic rhetoric, while defending itself by saying things like “truth ain’t always comfy.”

This “significant improvements” announcement came very shortly after Grok attacked Elon himself by saying things like he’s “the top misinformation spreader on X,” and that his Nazi salute at Trump’s inauguration was fascism. (44)

So, ya know…pretty interesting timeline of events there… 🤔

Loss of Desirable Difficulties

💡 Desirable difficulties are tasks or learning opportunities that require a level of effort to achieve, but upon achieving them, you feel a deeper sense of accomplishment.

Now, studies on AI’s effects on our brains are still young (because so is AI), but a study from MIT asked 54 participants, ages 18 to 39, to write several SAT essays using either ChatGPT, Google, or nothing at all. (42)

They found that the ChatGPT group had the lowest brain activity and “consistently underperformed at neural, linguistic, and behavioral levels,” as well as that over time, the ChatGPT users got lazier and lazier, using only copy-and-paste by their final essay.

The group that did not use ChatGPT or Google showed the highest levels of neural connectivity, were the most curious and engaged, and expressed higher satisfaction with their essays.

The Google search group also expressed high satisfaction and more active brain function than the ChatGPT group.

The brain is a muscle. It needs to be used to maintain its abilities.

But when we begin to rely on AI (and as we’ve seen, it’s a very slippery slope from dabbling to depending), we willingly begin to give away our brain exercise. We no longer need to read articles and synthesize information — we type in a question and out pops an answer. And over time, we begin to lose the ability to do the kind of complex thinking that requires patience and a willingness to be uncomfortable. (47)

“LLMs are designed to make cognitive work effortless, but that feels so icky because for it to be worthwhile at all, it has to be effortful.” (46)

 

“Research on folk concepts of the good life reveals that while people claim to prefer ease, they consistently rate lives involving effortful engagement as more desirable and morally superior. Whether drafting a manuscript, running a marathon, or raising a child, meaning comes from being able to attribute success to one's own effort in overcoming difficulty. This effort lends value to the work. More generally, the sorts of pursuits that people describe as meaningful are often those that they work the hardest on.” (103)

And then there’s yet another Threads post about AI from two weeks ago where someone — in full earnest — said, “Today I kinda realized how useless I am without AI. I was on a plane and the wifi went out so I couldn’t talk to my agents. Everything I wanted to do immediately felt exponentially cumbersome without AI assistance and I quickly gave up and took a nap.”

And when people in the comments tried to show him how concerning and dangerous that was, he doubled down about how much he appreciated AI, because he got to take a nap.

…What?

I’ll also say that I know in my soul that there are people who have fed this entire AI blog post series into an AI chatbot and asked it to summarize it for them. And I’m not gonna lie to you…that really bums me out. I don’t want anyone giving up their cognitive abilities! It’s why I include an audio version of [almost] every single blog post we do — to make it as easy as possible for you to consume and absorb the information.

So if you’ve actually read or listened to every part of this series, will you please comment below and let me know? I’d love to hear from you!

“We've always built tools that extend human ability. The hammer extends the hand, the car extends the foot, and the computer extends memory and calculation. Those tools changed how we live, but they didn't really fundamentally alter how we think—our sense of where agency resides. AI changes the game because it doesn't just support our thinking; it increasingly participates in the cognitive process. The issue isn't just about performance; it's psychological. When a person grows accustomed to reaching outward before reaching inward, the center of gravity in thought begins to move and perhaps even loses its essential anchor.” 

– John Nosta, Psychology Today (52)

Photo credit Andrea Gucklhorn via Unsplash.

The Commodification of Intelligence

Sam Altman literally wants to sell you intelligence.

Literally.

In his own words, he said: “We see a future where intelligence is a utility like electricity or water, and people buy it from us on a meter.” (53)

Even worse, he wants to sell it to you on a scale based on how much you’re willing to pay for it.

If you want to spend less on intelligence, you’re going to get less intelligent results. If you want to spend billions of dollars on intelligence, you’re going to get, for lack of a better term, the most intelligent results.

He literally said this. You can watch the video here.

So, just to reiterate here, AI is trained on stolen work.

Stolen art.

Stolen thought.

And Sam Altman wants to turn around and SELL US BACK OUR OWN INTELLIGENCE AFTER STEALING IT FROM US IN THE FIRST PLACE.

Meanwhile…the head of ChatGPT, Nick Turley, said that “the company’s mission is to keep this tech accessible to benefit everyone: ‘And one of the ways that we do that is by making sure people can access it for free.’” (54)

He then also tried to play it off like OpenAI wasn’t trying to make money, but then said, “If we do our job right, many people are going to want to pay and subscribe.” (54)

Lol okay dude. You’re not fooling anyone.

The “Homogenization of Language”

I came across this quote from Daphne Ippollito, natural language processing researcher at Carnegie Mellon University in the research process for this series and I can’t get it out of my head (45):

“Something I’ve seen happening at various levels in academia, and I’m sure outside of academia as well, is: when people start to rely on AI so much to help their writing, they start talking and communicating like the AI. So they will say things to communicate things that are authentically them, but their authentic self and how they communicate has been influenced by their usage of AI.” 

This is just so sad.

The way we speak, the way we communicate, the words we use are part of what makes us who we are. But using LLMs as much as we are is causing us to all start to sound alike.

I want no part in that.

No More Joy in Discovery

This is the one that takes the most wind out of my sails on a regular basis. (Pun incoming and intended…)

My main music playlist is every bit of my personal brand of nerdy heaven: bagpipes, sea shanties, hurdy gurdies, viking rock… Basically, anything that makes me feel like I’m going on a quest, mostly of the seafaring variety 🏴☠️

I get so. much. joy. from this playlist.

I started building and carefully curating this playlist back in 2021, and it’s nearly 10 hours long now.

Betcha didn’t think a 10-hour playlist of bagpipes and sea shanties was possible, didya? 😅

But this playlist has been stuck at about 10 hours long for over a year now.

Why?

Because I had to turn off “smart recommendations” for it, because almost overnight, every new suggestion was AI. I could no longer experience a song for the first time with joy and wonder, because every song became a research project into whether or not it was actually made by humans. It was deflating and exhausting.

And on a probably-much-more-relatable note… Last week my husband and I started watching Our Oceans on Netflix, which came out in 2024, so well before AI-generated videos were a thing, let alone convincing.

But by the end of the first episode, instead of being enthralled and captivated by the beauty and wonder that is our planet (which, yes, there were some of those emotions, too), I found myself angry. Because throughout the entirety of the episode, my husband and I had to keep reminding ourselves and each other that the footage we were seeing was real — not AI-generated.

We are living in a world where we literally don’t know what’s real anymore.

Not just what’s true. What’s REAL. And I really fkn loathe that.

Photo credit Bruce Timana via Unsplash.

Existential Threats and Fear Mongering for Profit

If your algorithms are anything even remotely close to mine, you’ve seen the posts and news articles this week about how AI researchers and developers are concerned that AI could kill us all in the next few years, and you’ve also likely seen the follow-up posts from Dario Armodei (the CEO of Anthropic), Sam Altman, and Elon Musk, calling for slowing down and regulating AI technology advancement.

Yes. This is scary. And yes. We should ABSOLUTELY be questioning everything, slowing way, way, way down, and regulating the ever-loving shit out of AI. Like, yesterday.

Yes and…

Let’s also think about this:

When has a corporation or a billionaire ever ONCE done something “good” out of the goodness of their own hearts? 

Go ahead. Name one. I’ll wait.

Exactly.

Now. These are just my own beliefs based on the EVERYTHING that I’ve outlined in this series and a whole lot more that didn’t make the cut, but… I’m not convinced that (at least on some level) this isn’t a coordinated effort to:

(A) Get the general public to feel like all hope is lost and that we should just give up fighting and advocating for a safe, healthy planet to live on,

(B) Make us believe that their AIs are more “intelligent” than they actually are so that we invest in their very-soon-to-go-public IPOs,

(C) Distract us from the very real harm that AI is doing right now,

(D) Lay the foundation for a future excuse to their investors as to why their technology isn’t progressing as fast as they promised and still isn’t making money (and has a 5-to-1 debt ratio that is unheard of in any other industry), and

(E) Convince us that they must continue their AI arms race, because if ~they~ don’t develop the technology then another company will, and that company will be ~bad~. But ~they’re~ the good guys, so they should keep going. 🙄

And, ironically, about 2 hours after I wrote that last line, this ‘lil screenshot came across my algo…kinda proving my point… But, ya know, that’s just me… 🤷♀️

Is it a cause or a symptom?

The more that I’ve thought about the ethical, moral, environmental implications of AI and why, as a society, so many people seem to be drawn to it, the more that I believe that a lot of these issues don’t actually stem from AI itself, but are rather symptoms of larger issues…all rooted in late-stage capitalism.

Do we REALLY need to spend billions of dollars to boost productivity by 2%? (I saw that stat in one of the 103 articles used for this series, but I don’t remember which one.)

Do we REALLY need to be churning out content for the sake of the algorithms?

Do we REALLY need to be doing any of this at all?

What if instead we just…didn’t?

What if we were more intentional with our time and our attention and the way we spend our lives? 

I know that these are bigger questions than I or any one person can answer, but if you know me even one iota you know that I love asking, “What if…?”

What if…things were done differently?

What if…we didn’t have to work to live?

What if…we didn’t base the very foundation of our economy on extracting as much labor and productivity from the working class just to boost the profits of a dozen conglomerates and their multi-billionaire CEOs?

What if…we didn’t destroy the planet for the sake of shareholder profits?

What if AI isn’t the cause of our problems, but just the most recent symptom of an entire system gone wrong?

What. If.

But hey just in case this wasn’t clear…just like the saying “there’s no ethical consumption under capitalism,” this isn’t a green light to shrug our shoulders and say, “Oh well, there’s nothing I can do so I guess I’ll just keep doing what I’m doing.” Knock that off right now, friend — you’re better than that 💙

Okay so what can we do?

And now, for the part you’ve been patiently waiting a whole month for… 🥁🥁🥁 What on earth can we actually ~do~ about all of this?? Because despite what the tech bros and AI CEOs want you to think, there’s a lot we can do.

• Reject the idea that AI is an inevitability. Because if renewable energy, tree planting, and imprisoning for life all of the pedophiles running our country isn’t an inevitability, then why should AI be?
  
  

• Stop using generative AI. ESPECIALLY for things that give up your own agency, creativity, and mental capacities. The entirety of the universe has successfully existed without AI until 4 years ago. You’ll be fine, I promise.

And on that note, something I’ve been thinking about a lot over the last year or so that I haven’t quite fully fleshed out my thoughts on yet is the cost of convenience. What’s the true cost (in every sense of the word – monetary cost, environmental cost, cognitive cost) of convenience?

There’s also a quote that I read early in my Master’s program that has stuck with me ever since:

“Faced with a conflict between a desire for self-gratification through unsustainable behavior and the knowledge that the environment is threatened by such behavior, we repress our awareness of the conflict, deny the threats that face us, displace them onto other communities, and rationalize our continued unsustainable behavior as having no alternative. Denial and other defensive thinking are particularly likely when people believe they can do nothing to lessen the danger.”

So if you use LLMs and your argument in favor of them is that they’re more “efficient” than Googling something, with the above quote in mind, I want to ask you, without judgement, without shame…Is the convenience worth it?

Is your one query worth using 100x more energy than a Google search?
Is it worth draining communities of fresh drinking water?
Is it worth the risk of your answer being incorrect more than half the time?
Is it worth all of the stolen writing, art, music, code that built the LLM?
Is it worth outsourcing your own cognition and what makes you you?
Is it worth supporting billionaires who want to sell our own stolen intelligence back to us?

Is it worth it?

I can’t answer that for you. But I want you to be honest with your answer.

And remember…we can do hard things.

• Turn off AI features as much as you can, everywhere you can.
  
  

• Add “-ai” to your searches. It’ll remove the AI summary at the top of the page.
  
  

• Switch your internet browser to one that’s AI-free and protects your privacy. Vivaldi and Waterfox are the two that seem to be the most popular. I’m in the process of testing out / switching over to Vivaldi, and my husband has already switched to Waterfox without complaint.
  
  

• Switch your default search engine to one that doesn’t use AI. DuckDuckGo has a no-AI option, or you can easily turn off all AI features in the regular DuckDuckGo, though apparently every time you delete your cookies they turn back on. Ecosia is another one that makes it super easy to turn off AI features, and they also plant trees and donate to climate action with the profits from ads.
  
  In transparency, I used Ecosia many many years ago and found the results, unfortunately, a bit wanting at the time, and went back to Google. My husband switched his default search engine to Ecosia when he switched his browser to Waterfox, and did say that it’s “less than great,” but it doesn’t sound like it’s been too much trouble for him. I’ve set my default search engine to Ecosia, and we’ll see how it goes 🤞 If I don’t love it, especially for doing blog post research, I’ll give DuckDuckGo a try.
  
  

• Attend town halls in your community. Fight for your local officials to pass data center moratoriums, or better yet, outright bans.

My parents are currently in the process of fighting two data centers in their community, including one that’s proposed location is essentially in their backyard, and the other just two miles away. I asked for their suggestions based on their experience so far, and my dad responded:

“I would say it’s easier to stop one before the process starts. Once the train leaves the station it has momentum that’s hard to slow down. If your city hasn’t already passed rules against one, push them to pass them. Find out if any of your city leaders have signed NDAs. Watch for requests for zoning changes. Ask candidates running for office in your community if they support or oppose data centers.”

💡 If you’re in Pierce County, the county currently has a survey open to share your thoughts and opinions about data centers in the area and future data center regulation. Go fill it out.

• Call your representatives. Make it known that you want to see real action to implement safeguards against AI and data centers. Our president isn’t gonna do it. His goal is to make AI move faster and remove any friction for these AI companies to grow without constraints. And with his response to the CEOs of the top AI companies saying there should be regulations and guardrails in place being a resounding, “absolutely not,” we gotta fight it at every other level of government.
  
  

• Talk about it! Share this series with others. Talk about the impact that AI is having on the planet, on communities, on our own brains. Talk about why it matters to you.
  
  

• VOTE. For the love of everything and anything, please vote. At every level of government. Midterms are coming in less than two months. Find out what your candidates’ views are on AI, who’s financially backing them, and vote like your life depends on it.

And most importantly, I need you to hear me when I say this:

Know that we are not cooked. We are not f***ed. 

Yes, things are scary. I won’t lie to you and pretend that I’m not scared, too.

But the future is not written.

These tech bros and billionaires and fossil fuel companies want us to believe that it is so that we stop fighting.

We. Will. Not. Stop. Fighting.

Like I said last week, we cannot separate individual actions from collective impacts. Every single one of us has the ability to do SOMETHING.

So stop using genAI, talk about why you aren’t using it with the people around you, and no matter what, get out and VOTE on (or before) November 3.

 

💡 View all of our sources used for this blog post series here.