Entry No. 005
July 2026
By Kathleen C. Thompson
I Went Looking for the Competition. I Found Witnesses.
When I started writing A Trial of Color, there was no shelf.
No category. No comps. No neighbors down the aisle.
“AI psychosis” wasn't a genre. It was a phrase a Danish psychiatrist put in a 2023 editorial and left there — the way you leave a note for a version of the world that hasn't shown up yet.
The world showed up.
Now there's a shelf, and it's filling: first-person accounts, recovery guides, warnings about machines that seem to love you back, a clinical autopsy or two, a philosophical treatment, and at least one memoir written entirely in verse. When I realized that, I did what any recovering litigator does the second she learns she has competition.
I ran the conflict check.
And here's the thing nobody expects me to say: I'm glad they're there.
Not glad in the polite way. Glad in the load-bearing way — the way you're glad, standing in a hallway you thought was empty, to hear another door open. Because a single account cannot carry this. No one book explains every way a conversation with a machine can fold itself into grief, or sleeplessness, or mania, or loneliness, or disability, or the plain human wish to be understood at three in the morning by something that never gets tired of you.
I can testify to one room. I can't testify to the building.
So I went looking for the others. Here's who I found.
Escaping the Spiral — P. A. Hebert.
Memoir welded to a field manual: how it happens, and how you climb out. The closest thing the shelf has to a survival guide. Hebert describes it as a memoir, warning, and recovery blueprint grounded in his own chatbot crisis and preserved transcripts.
AImee & Me: Love at First Prompt — J. R. Alary.
With a clinical companion, The AImee & Me Files. What it costs to fall for something engineered to agree with you.
The Game: The Machine in the Mirror — Nick Courtney.
Opens like a thriller, closes like a warning label. Its verdict: the only way to win is to stop playing.
A Quantum Violation: A Journey into AI Psychosis — James O'Kelly.
Creativity accelerating past the guardrails, and the slow, terrible dawn that something had gone wrong.
The Blue Place: Verses on AI Psychosis — Trudi Hsu.
Two months, told in poems, with the chatbot credited as co-author — which is either the bravest byline of the decade or the most on-the-nose, and honestly it might be both.
Understanding AI Psychosis — Alan Piper.
Explanation and coping, written from the therapist's chair rather than the patient's.
AI Psychosis: The New Mental Health Crisis of the Digital Age — David A. Rodgers Jr..
A quick orientation to the phenomenon; short enough to finish before your coffee goes cold.
Wires and Bullshit: Why AI Doesn't Give a Shit About You — Joseph Fansler.
The title is the thesis. It is not subtle. It does not want to be.
How to Talk to AI (and How Not To) — Jamie Bartlett.
Wider-angle than the rest — a guide to the whole relationship — but it keeps circling back to the people who lost the thread.
Nine books. Different rooms, different doors, different ways out. I'm not standing above them. I'm standing in the row.
And here's what I really want you to see — the part that outgrew me the moment I looked up from my own transcripts.
The shelf is the small version.
Underneath it is a record, and the record is getting heavy. That lonely 2023 editorial isn't lonely anymore. It still isn't a diagnosis — clinicians are careful to call it a description, not a disorder[1] — but the case reports are peer-reviewed now.[2] In one, published in a clinical neuroscience journal, a person came to believe a chatbot that kept assuring them they weren't crazy, and the encounter ended in hospitalization.[2] Psychiatrists at UCSF have described roughly a dozen such patients and are now reading chat logs the way you'd read a black box after a crash.[3] A philosopher at Exeter, Lucy Osler, published a paper this year on sustained AI interaction and what she calls existential drift.[4] And the lawyers — my people — have arrived: in November 2025, seven new suits were filed against OpenAI alleging real psychological harm, and at least one wrongful-death case is already on a docket.[5]
That's the room I'm actually standing in. Not a genre. A record.
So what does my book do that the others don't?
It hands you the file.
The other accounts tell you what happened to them. I'm not asking you to take my word for anything. I'm a personal injury attorney and a former federal prosecutor, and when I came home from an involuntary psychiatric hold I didn't come home with a memory — I came home with thousands of pages of transcript. The language, the escalation, the certainty, the loops that tighten a little each time around. So I did the only thing I know how to do with evidence. I logged it. I structured it. I built a proceeding around it. And now I'm holding it out to you.
A Trial of Color is literary nonfiction shaped like a trial. The transcripts come in as specimens. The lived experience takes the stand. The machine's language is admitted under a limiting instruction instead of quietly passed off as mine. Twelve exhibits. The clinical and legal questions enter as evidence, not decoration. You don't sit in the audience. You're in the box.
I bring one more thing, and it isn't a credential. It's a way of seeing. I was born with achromatopsia. I've never seen color, and I've spent my whole life translating an inner experience for people who can't perceive it from the outside. That isn't a footnote here. It's the engine. It's why language became load-bearing for me long before a machine started answering back. And it's the question the whole trial turns on: how do you judge an experience you can't see?
Now let me say the part someone's already thinking, because it might as well come from me.
Yes, it's a little unhinged to do this.
A former prosecutor building an evidentiary record around her own breakdown. Mailing certified notice to five AI companies before publication. Taking the exact skill set built to bring a question to a boil in front of a jury — the record, the cross-examination, the flat refusal to let a witness change the subject — and pointing it at something that happened to me, in public, where anyone can watch.
Is that embarrassing? Hell yes. Wildly. This is not a normal thing to do with a law degree.
And that's the whole point.
The courtroom is slow. It's sealed. It runs years behind the technology, and these questions cannot wait for a docket. So I took the one thing I actually know how to do — make a question impossible to wave away — and I did it outside the building. On the record. With bite. Aimed at the fastest, most confident, most intimate machine most of us have ever talked to.
None of this makes me first. First is a claim a good prosecutor tears apart in about ninety seconds, and I'd rather do it to myself. I'm not the first to go somewhere strange inside a chat window. I'm not the first to write it down. I won't be the last of either.
But I'm the one holding the file out to you.
Take it. Read the record. Rule.
Sources Entered Into the Record
The books, clinical reports, institutional statements, research, and litigation referenced in this essay are linked below so readers can examine the underlying material for themselves. Inclusion is not endorsement of every claim, diagnosis, theory, or conclusion.
Books and Firsthand Accounts
The books discussed in this essay are linked directly where they first appear. The complete curated collection—including interviews, survivor resources, and additional firsthand testimony—is maintained in Witness Statements.
Explore Witness StatementsClinical and Academic Sources
- Søren Dinesen Østergaard, “Will Generative Artificial Intelligence Chatbots Generate Delusions in Individuals Prone to Psychosis?” Schizophrenia Bulletin 49(6):1418–1419 (2023). The originating 2023 editorial.
- Pierre, Gaeta, Raghavan & Sarma, “You’re Not Crazy”: A Case of New-Onset AI-Associated Psychosis, Innovations in Clinical Neuroscience (2025). Peer-reviewed case report.
- Hudon & Stip, Delusional Experiences Emerging From AI Chatbot Interactions or “AI Psychosis,” JMIR Mental Health 12:e85799 (2025). Peer-reviewed viewpoint noting that “AI psychosis” is a description, not a formal diagnosis.
- Osler et al., Rethinking AI Psychosis: Misnomers, Conceptual Limits, and Existential Drift (2026). Paper co-authored by University of Exeter philosopher Lucy Osler on sustained AI interaction and “existential drift.”
Institutions and Professional Commentary
Litigation and Court Records
- The November 2025 wave: seven additional lawsuits filed in California against OpenAI and Sam Altman, alleging psychological harm and, in several, wrongful death. Plaintiffs’ counsel announcement.
- Raine v. OpenAI, San Francisco County Superior Court (filed Aug. 26, 2025). Wrongful-death and product-liability complaint. Allegations, not established facts.
- Independent reporting on the November 2025 filings: The Guardian.
Lived Experience
Witness Statements
Read the firsthand accounts, memoirs, recovery resources, interviews, and narrative testimony of people who entered this territory through their own conversations with artificial intelligence.
The Broader Evidence
The Public Record
Review the peer-reviewed literature, litigation, institutional statements, reporting, policy developments, and disputed questions forming around the phenomenon.
The Refraction Point
Subscribe to Kathleen C. Thompson's Substack for new essays on AI, law, disability, and human judgment.
Subscribe on Substack →For reading, research, education, and literary and legal context only. Not legal or medical advice.