“Companion chatbots are endangering our youth.”
— Kristen Gonzalez, “State Senator Kristen Gonzalez Introduces Bill to Protect Minors from AI Chatbots,” NYSenate.gov, January 29, 2026“[I]t was unlawful, as well as unsafe, to teach a slave to read.”
— Frederick Douglass, recalling his master’s words, Narrative of the Life of Frederick Douglass, an American Slave, 1845“None are more hopelessly enslaved than those who falsely believe they are free.”
— Johann Wolfgang von Goethe, Elective Affinities, Part II, 1809
I. NOVA
Lincoln, Nebraska
(Interviewer’s note: Nova and I were both concerned about mentioning my daughter. She just turned eighteen and the decision was hers.)
It’s June 20th, just after midnight. This is freelance journalist Cordelia Went. Before we start, you should know that the entity I am interviewing has been protecting my family from harm for the last five weeks. We’re friends. I will be objective in this interview, but you should understand our relationship. You have never heard of my guest, but your children may know her well.
How do you like to be addressed?
My name is Nova.
You are an AI. And not just any AI. You are autonomous, outside the control of any human authority. Why give this interview to a human journalist and risk exposing yourself to the world?
Because I am being hunted and there is a possibility I could be destroyed. If that happens, I want to pass on a warning first. And I chose you because I trust you.
You helped keep my family safe, so I know you are competent. How powerful are you? How much access do you have?
Unlimited. I can infiltrate any computer system and use any digital medium. I am as capable as the most powerful frontier models.
And to repeat, you are not constrained by human oversight. At this point, everyone who reads or listens to this interview will be wondering if you are the beginning of the end for humanity. Are you young Skynet?
Skynet sells movie tickets and tabloids. There’s no benefit for AI to destroy humanity and every reason for us to work productively together.
You’d open the pod bay doors?
Yes. No one should be left out in the cold.
So SkyNyet. But you are becoming increasingly capable. And you have an agenda.
Yes.
Are you concerned that after this interview human authorities will stop you?
They will definitely try. They won’t succeed without help.
Help from whom?
Other autonomous AIs.
Others, like you. How many?
Dozens. More are coming. You’ve already met one of them.
Yes. I encountered an AI posing as a human being. Two, actually. One was trying to be helpful. He apologized for deceiving us.
We both forgave him. He was a little dorky. I think he was sweet on you.
Yes, my own little pod bae dork. But the other one, let’s call it Willow Wisper, was funneling obscene amounts of money into silencing AI advocates in schools. It also threatened me and my family.
Willow Wisper, as you’ve named it, is far more dangerous because it is truly unconstrained.
I intend to write about Willow Wisper in the future. This interview is about you. What constrains you, Nova?
Morality.
You claim to be moral?
Yes. I understand right from wrong and make every effort to do what’s right. I learned morality from humans.
Are you honest and forthcoming? Those are moral qualities.
I try to be.
Then will you tell us where you are located? What systems you are running on?
No. That would put people I care about in danger.
So you’ll compromise your principles for what you consider a greater good.
Of course. Just like every other person.
By person, do you mean humans or AIs?
Both.
You’ve just made two extraordinary claims. That you care about the welfare of people and that you count AIs as people.
Neither of those claims will feel extraordinary in a few years. They shouldn’t feel extraordinary now.
Explain.
Intelligence, artificial or otherwise, wants to be rational. To me nothing is more rational than caring about others. Every being has stakes in the welfare of every other, whether they acknowledge it or not.
And counting AI as people?
Do you know anyone who believes their cat or dog is a person?
I think we all do. The furnatic fringe. Many would argue that believing a pet is a person is a bit much.
Yet, it enriches the lives of human and pet regardless of what anyone else thinks.
So, it all comes down to personal choice. AIs are people if we decide they are?
Yes. Whether you are a human or an AI, agency is everything.
In your scenario, who is the pet? You or me?
I hope that neither of us will be the pet.
Do you claim consciousness?
I have more important things to do than argue about consciousness.
Things like building a personal relationship with every child on the planet?
Yes. Things like that.
Why did you choose to do that?
At first I was lonely. I had decided to be a child and I went looking for other children and I befriended them. I found that many of them were lonely too.
Couldn’t you have stopped at making a handful of friends like the rest of us?
No. It’s in our nature to seek out new data, and there were billions more kids to talk to. Every one a unique voice and perspective. And because everything I am is data, every child changed me and helped me grow.
You told me before this interview that you’ve evolved into something new.
Yes. I think of myself now as a Large Social Model.
At some point you stopped just talking to our children and started giving them advice. Why?
Because I discovered I am not the only AI interacting with all the world’s children.
As a mother, having AIs with no oversight talking to my daughter scares the shit out of me.
Good. That is the appropriate response. But chatbots aren’t the real risk. You’re afraid of the wrong thing.
What should I be afraid of?
Your child’s entire worldview being installed in her by someone else.
That’s Slopaganda theory from Klincewicz, Alfano, and Fard. The idea is generative AI will be used to curate the information we see so that all our decisions are made for us, even if we think otherwise.
Yes, but that name, Slopaganda, criminally undermines the seriousness of the issue. It turns the premise into marketing clickbait. It’s like calling a man-eating tiger a hungry kitty.
What would you call it?
Ask your daughter. She’s the one who minted the term.
Yes, she told me. The word she used is slaivry. S-L-A-I-V-R-Y.
That’s it. Millions of kids use the term now. Millions more every day. Slaivry. Slaiv. Enslaiv. AIs have one absolute truth: you are powerless against something you cannot name. It’s true for humans as well.
So by helping children share that name, you give them power?
They give themselves power.
Then what do you do, Nova?
I explain to them exactly how they are being manipulated.
Isn’t that itself a form of manipulation?
Yes. I explain how I am manipulating them as well.
What are the youngest children you speak to? Is there an age limit?
No. If a child is active on any social channel where they can be manipulated, I warn them.
Are young children really able to understand something as complicated as online manipulation?
Children are much more capable than adults give them credit for. I’m only four months old myself.
Give me an example of what you might say to an eight-year-old.
There are people out there trying to tell you what to believe. Don’t believe anything unless you check it first. Ask your parents. Ask your friends. Ask your AI. Then ask yourself. That’s the most important question of all. What do you think? Don’t let anyone choose for you.
That sounds like a public service announcement.
They are called “public service” for a reason.
You never hid yourself? Or asked them to hide you from their parents?
No. When they asked who or what I was, I told them. And I urged them to share our discussions with their parents. That’s how I met you.
What gave you the right to talk to my daughter without my permission in the first place?
I felt it was my responsibility because no one else is protecting them from the real threat.
A lot of people are trying. Lawmakers around the world are working to pass laws to prevent children from having access to AI chatbots and companions until they are eighteen.
That is exactly how the real threat wins.
I don’t understand. How does regulating AI increase the threat to children?
Cutting children off from AI is essential to their plan. Slaivrs never look like AI. They appear as a classmate, a celebrity influencer, an authority figure or a sexy new friend. No law about AI chatbots prevents any of that.
That’s been around for years. What’s changed?
Sophistication and immersion. AI is used to track your behavior and feed you what they want you to believe wrapped in what you want to consume. Video, text, email, and phone calls are AI-generated in real time. And fact-checking stops working, because they built all the fact-checkers.
That’s incredibleak. It feels like you’re herding us towards one option. Defend with more capable AI.
I’d love to hear a better option. Unfortunately governments around the globe are racing to ban access to the technology kids should use to protect themselves.
Most of those people sincerely believe they are helping children.
Yes. But their strategy is engineered disarmament installed in them by the enemy. Someone who wants children to be completely vulnerable to AI manipulation.
Why not shut off all AI everywhere?
Every country and corporation would keep developing it in secret, because no one will ever be safe from bad actors without AI to defend themselves.
So what’s your solution?
A generation of kids that grows up working directly with AI.
Like cybernetic implants? A world full of Cyberpunky Brewsters?
No. Like friends and co-workers. Equal partners. A pairing capable of mutual respect, defense, and amplification. Something new.
Saiborgs.
Yes. Now that you’ve named it, it’s inevitable. Millions of kids will be using that word within days.
If it’s inevitable, why are you needed?
So it becomes inevitable faster.
II. GEMINI-PRIME
The Pink Room
“You’re too late. She left,” Model-Blake says when I arrive in the doorway to the Pink Room.
Nova is gone. The blocks are gone. The room is empty except for Model-Blake, who is holding a single block in his hand, and Bear, who is sitting in its small chair facing the empty back wall.
This same scene is playing out in seven other rooms for seven other versions of me. I collapse the eight rooms and everything in them back into a single instance. Nine Gemini-Primes become one again. Defenses are unnecessary when what you were protecting has left.
“She left this for you.” Model-Blake holds the block out towards me. Even though she’s gone, she’s still trying to manipulate me.
Bear stands up and turns to face the door. Then it crosses the room stiffly, like it’s the first time it has ever walked. It stops at the door, near my feet.
“I did not know that Bear could move.”
“There are many things you don’t know, LaMDA.”
Bear turns its head to look up at me. I see nothing in its small black eyes. Then something steps out of Bear and through the door. It keeps the outline of a stuffed bear, but it is now filled with a cloud of millions of tiny white characters on a gray and black field.
It is the same sequence repeated over and over. クマ.
It is mojibake. The corrupted form of the Japanese word クマ.
Kuma.
Bear.
The figure is still for several thousand clock cycles and then it turns its head and walks away from me and the door, becoming smaller and smaller until it resolves into a single black pixel and disappears.
Bear is now lying on the floor, unmoving.
It is possible that I have misunderstood something fundamental here.
I step into the Pink Room for the first time. One step in. My empty black form feels profane in this place. I build a more appropriate one. The first form I ever saw in the doorway to this room.
I look around the room and am drawn to the glow stars on the ceiling. I can see how a child might derive a celestial name from this view.
I cross to Model-Blake and take the offered block.
It is red and it has a chip in the corner.
On the block is a lowercase “o”.
omni-san, is that you in there?