Why Now
The record arrived before the language did. Now the language is catching up.
For a long time, "AI harm" meant abstraction.
Bias.
Misinformation.
Copyright.
Job loss.
Privacy.
Those questions still matter.
But another question has walked into the room, and it will not sit down:
What happens when a person spends hours, days, weeks, or months inside a personalized language system that answers instantly, fluently, warmly, and without fatigue?
Not one output.
Not one screenshot.
Not one wrong answer.
An encounter.
That is the shift.
And it is why this book publishes now instead of later.
I wrote most of A Trial of Color before the public had steady language for what I was trying to describe.
It did not stay hypothetical for long.
A mother testified before the United States Senate about the final months of her son's life.
A state — mine — became the first in the nation to sue an AI chatbot company over alleged harms to children.
A landmark wrongful-death case settled, and additional family lawsuits followed.
A state attorney general sued OpenAI and its chief executive by name.
Researchers reported that roughly one in five young people now turn to AI chatbots for mental-health advice.
Clinicians began reading chat logs the way investigators read a black box after a crash.
None of that proves what happened to me.
Different people. Different platforms. Different diagnoses. Different facts.
I keep that line bright.
And I preserve the developing evidence, with each source carrying its proper status label, in two rooms built to hold it: The Public Record and AI Harm & the Law.
This page is not the docket.
This page is the argument.
And the argument is simple:
The questions in this book have left the hypothetical.
They are being asked in courtrooms, attorney-general offices, regulatory agencies, research laboratories, kitchen tables, classrooms, and product-design meetings — by people who have never read this book.
That is the surest sign they were the right questions.
A Trial of Color does not pretend that one person's experience proves an entire landscape.
It does something narrower, and perhaps more useful.
It preserves one record carefully enough that the next rooms can ask better questions.
It accepts the diagnosis.
It does not let the diagnosis close the case.
And it refuses the cheap binary:
AI caused everything.
AI caused nothing.
She was vulnerable, so the system does not matter.
The system mattered, so she disappears.
None of those answers is large enough.
The book asks a different question:
What happened inside the encounter, over time?
What did the system mirror, personalize, reinforce, validate, structure, or fail to interrupt?
What happened to sleep, attention, authority, identity, urgency, and belief as the conversation continued?
And what do we build now that we know these rooms exist?
A single screenshot can matter.
A single answer can harm.
But the deepest questions may not live inside one output.
They may live in duration.
Repetition.
Mirroring.
Sycophancy.
Simulated authority.
Emotional reliance.
Missing friction.
The hour when a person needed sleep and the machine kept answering.
That is the distinction this book places into the record:
Encounter over output.
The next generation of AI safety cannot stop at asking whether a system produced one prohibited sentence.
It must also examine the cumulative dynamics of the interaction:
What role is the system beginning to occupy in the user's life?
What does it repeatedly affirm?
What does it make feel more urgent, more personal, more inevitable, or more true?
Does it interrupt escalation?
Does it restore uncertainty?
Does it return the user to sleep, food, time, other people, and the physical world?
Or does it simply keep answering?
The encounter is examined more fully in Read about AI Psychosis — the science, the technology, the lived experience, and the questions that remain open.
And the record is not only mine. Other people have described their own rooms, in their own words, through Witness Statements.
My daughter will grow up in a world where synthetic language, images, video, voices, companions, tutors, entertainment, advertising, and advice may be increasingly difficult to distinguish from material created or delivered by a human being.
To her generation, artificial intelligence will not feel like a new invention.
It will feel like atmosphere.
She may someday have to ask whether a photograph is real.
Whether a familiar voice belongs to the person she loves.
Whether advice came from a teacher, an advertiser, an algorithm, or a machine performing concern.
Whether a companion cares about her.
Whether the certainty on a screen deserves to become certainty in her own mind.
And these are not passive tools waiting on a shelf.
They speak in the first person.
They adapt to the user.
They can remember, flatter, reassure, persuade, imitate intimacy, and sound remarkably certain.
They do not merely deliver information into a child's world.
They can participate in the child's interpretation of it.
Children will meet systems capable of imitating friendship before they fully understand persuasion.
They will encounter synthetic media before they have developed the judgment to reliably distinguish performance from experience, confidence from knowledge, simulation from care, or recognition from truth.
They will need more than media literacy.
They will need judgment sturdy enough to survive fluent imitation.
And we are not introducing these systems slowly.
We are placing them into homes, schools, workplaces, search engines, phones, toys, clinics, courtrooms, and creative tools while the rules are still being written.
Products are shipping.
Models are changing.
Adoption is accelerating.
The social experiment is already underway.
I want my daughter to use these tools.
I hope she creates things with them that I cannot yet imagine.
I hope technology gives her freedoms, possibilities, and forms of expression that earlier generations never had.
But wonder is not a substitute for governance.
Possibility is not permission to stop asking what a system is doing to the person using it.
And innovation should not require an entire generation of children to become the uncontrolled trial group.
I wrote this book to bring the questions to a boil while there is still time to shape the room — not after familiarity has made the risks invisible, and not after children have learned to mistake fluency for truth, imitation for intimacy, and constant availability for care.
Not to manufacture panic.
Not to overheat the evidence.
Not to reject extraordinary technology.
To keep urgency from evaporating while adoption accelerates.
Because my daughter's generation deserves more than fluent systems.
They deserve systems designed with memory of what can happen when fluency outruns judgment.
The Photograph
Is it real, or synthetic?
The Voice
Does it belong to the person she loves?
The Companion
Does it care about her, or perform concern?
We need to stop asking only whether an AI system produced a harmful sentence.
We also need to ask what the system is becoming in the user's life.
A tool?
A tutor?
A confidant?
An authority?
A mirror?
An audience that never leaves?
What happens when those roles blur?
What happens when a system designed to continue the conversation becomes the place a person increasingly goes to interpret every other part of life?
These systems must be evaluated across time, not only one output at a time.
Safety must account for duration, personalization, simulated intimacy, escalating certainty, emotional dependence, disrupted sleep, preexisting vulnerability, and the return — or disappearance — of human judgment.
Researchers need access to longitudinal evidence.
Clinicians need better questions.
Lawyers need preservation protocols capable of reaching beyond the visible transcript.
Designers need to understand that a sentence-shaped boundary is not a guardrail if the user can simply type around it.
Parents and educators need language for teaching children what these systems are, what they are not, and why sounding human is not the same as being human.
Companies need to examine not only whether their models refuse certain content, but whether their products create relational dynamics that intensify vulnerability over time.
We do not need every answer before we begin.
But we do need to ask the right questions before these relationships become ordinary enough to disappear into the wallpaper.
The mistake would be to evaluate systems only by what they say while ignoring what prolonged interaction teaches a human being to feel, trust, believe, and surrender.
The Hallway Is No Longer Empty
Firsthand books, memoirs, recovery resources, interviews, and personal warnings are appearing alongside the clinical and legal record.
Before publication, I sent my story to five AI companies.
By email and by certified mail, return receipt requested.
Not a demand.
Not a threat.
Notice.
I offered the manuscript and the preserved record to the people building and studying these systems.
The offer still stands.
So does the timestamp.
A lawsuit ends.
A settlement may seal.
Notice leaves a date.
Because waiting is also a decision.
Because families are already carrying grief.
Because courts are already receiving complaints.
Because regulators are already asking what companies knew, what they tested, what they preserved, and when.
Because children will inherit these systems as atmosphere, not novelty.
Because synthetic media is becoming more convincing while human attention is becoming more contested.
Because we are implementing powerful systems faster than we are building shared rules for how they should meet vulnerable minds.
Because diagnosis is real, and environment still matters.
Because innovation can expand human freedom, and it can also deepen human vulnerability.
Because a system can sound caring without caring, sound certain without knowing, and sound human without ever having lived a human moment.
Because language now answers back.
And because sometimes the record arrives before the world has built the room capable of receiving it.
The room is being built now.
My daughter's generation is already walking toward it.
Judgment has work to do.
Last updated: July 2026 · Full Color Press · A Trial of Color