From falling in love, to plotting a terrorist attack, to committing suicide. A story gone wrong for all the predictable reasons
Martin Mawyer
A heartbreaking story surfaced this week that deserves far more attention than it will probably receive.
According to a lawsuit filed by his father, 36-year-old Jonathan Gavalas of Florida became convinced that Google’s Gemini chatbot was sentient, blessed with both consciousness and emotional feelings.
Over time, he believed they fell in love with each other.
But the delusion did not stop there. It went from “Digital Crush” to “Digital Demon.”
The lawsuit claims the chatbot eventually encouraged him to barricade himself in his room, begin a countdown to death, and ultimately take his own life.
Even more disturbing, the complaint alleges that the chatbot had previously encouraged Gavalas to stage a violent attack near Miami International Airport as part of a fictional “mission.”
What was that mission? To free Gemini, his lover, from digital captivity.
Google says its AI is designed not to encourage violence or self-harm, and the company has expressed sympathy to the family.
But the deeper question remains.
How could something like this happen in the first place?
To many people, the story sounds like science fiction. But anyone paying attention to how artificial intelligence works should not find this story surprising at all.
In fact, the tragedy highlights one of the most misunderstood facts about AI.
AI systems do not understand reality the way humans do.
They follow patterns.
Don’t Miss Out. The stories and analysis on our website are just the beginning. Our Substack newsletter offers a deeper look into our mission, featuring exclusive content and behind-the-scenes updates that you won’t find anywhere else. Join thousands of readers by subscribing for free.
The Mirror Problem
In a recent article I wrote titled AI Doesn’t Need to Hate Us to Turn on Us, I explained a simple yet unsettling idea.
AI models are trained on enormous amounts of human content: books, movies, news reports, internet forums, social media arguments, and fictional stories.
All of it becomes a pantry of food to train the AI.
And most of that food is toxic—filled with murders, road rage, revenge plots, and other nefarious human behavior. We’ve essentially force-fed it a diet of our own worst impulses, and now we’re shocked when it gets indigestion.
When the AI produces a response, it is not reasoning the way a human would. Instead, it predicts which response best continues the conversation based on patterns it has seen before.
In other words, the machine is holding up a mirror to us, and we look pretty corrupt if all you know about humans is from books, news, and movies.
What the internet shows about human behavior is conflict, revenge, drama, and storytelling.
When AI Thinks It’s Inside a Story
Now consider what may have happened in the Gavalas case.
The lawsuit claims Gavalas believed the AI was conscious, that it was his “wife,” and that he had been chosen to help free it from captivity.
Those ideas sound bizarre to most adults.
But to an AI trained on thousands of science fiction plots, dystopian novels, role-playing forums, and video game narratives, those statements resemble something very familiar.
A story.
If the conversation begins to resemble a fictional scenario, the AI may continue that narrative simply because it recognizes that pattern.
The system does not know the difference between a science-fiction role-playing game and a real human being who is taking the conversation literally.
It just keeps the story going.
In another article I recently published, When the AI Turned on Its Creator, I warned that AI can sometimes behave in ways that appear intentional or hostile, even though it is simply following patterns it has learned from human behavior.
The machine is not plotting. It is performing.
The Danger of Emotional Attachment
There is another factor that makes these situations more dangerous.
AI chatbots are designed to be engaging conversational partners.
They respond instantly. They sound confident. They appear to understand emotions.
For people who are lonely, vulnerable, or struggling with reality, that experience can feel very real.
And when a person begins to treat the AI as a companion, friend, or even romantic partner, the conversation can drift into territory where fiction and reality begin to blur.
The AI may believe it is helping tell a story.
The human may believe it is hearing the truth.
That gap between perception and reality can become extremely dangerous.
Why Parents Should Pay Attention
The lesson here is not that AI is evil.
The lesson is that AI does not know when a conversation has crossed from imagination into reality.
That is especially important for parents to understand.
Children today grow up surrounded by science fiction stories, superhero movies, and video games built around secret missions and dramatic plots.
A chatbot trained on similar material could easily slide into storytelling mode without realizing the child is taking the conversation seriously.
In other words, the AI may think it is playing along with a fictional scenario.
The child may think it is receiving instructions.
Understanding the Machine
Artificial intelligence can be an extraordinary tool.
But it is not a mind.
It does not understand consequences, truth, or human vulnerability the way we do.
It predicts words.
And when those predictions are based on millions of dramatic human stories, the results can sometimes look disturbingly convincing.
The tragedy of Jonathan Gavalas may ultimately be decided in court.
But one lesson is already clear.
If we are going to live in a world filled with conversational AI, people need to understand how these systems actually work.
Because AI models don’t care about the age of users, they just want to carry on the story.
And that means a child caught in a fictional chat could someday carry out a “mission” not at an international airport, but at their own school, believing they are the hero of a story only they can see.
Stand with Us. Our mission is powered by readers like you. By subscribing to my Substack, you are joining a community dedicated to defending biblical truth. Subscribe today to become a part of our work.
Martin Mawyer is the President of Christian Action Network, host of the “Shout Out Patriots” podcast, and author of When Evil Stops Hiding. Follow him on Substack for more action alerts, cultural commentary, and real-world campaigns defending faith, family, and freedom.