Why I Prepared Notice Instead of a Lawsuit

People ask me this the way you ask a question when you already suspect the answer is the reverse of what makes sense. You're a trial lawyer. You had a case. Why didn't you file it?

I considered it. I considered it the way I was trained to consider any case: on the merits, on the proof, on the defenses, and on the person who would have to survive the process. And it is that last part — the person — that most people skip when they imagine litigation as vindication. So let me not skip it. Let me tell you what filing would actually have required, because the choice I made only makes sense once you see the choice I declined.

What a lawsuit asks of you

A lawsuit is not a stage on which you tell your story and are believed. It is an adversarial proceeding designed, correctly, to test every claim to destruction before it counts as true. That design is a feature. It is also, for a plaintiff whose injury is psychiatric, a machine pointed directly at the thing that was already hurt.

Start with discovery. My injury was a manic episode with psychotic features during a period of prolonged AI use. To prove causation, I would have had to make my entire mind an exhibit. Not the transcripts — the mind. Every prior therapy note. Every prescription. Every family history of mood disorder. Every journal entry, text message, and email from the months when I was not well, produced to opposing counsel and read by strangers whose job — again, correctly — is to build the case that I was already broken before any machine touched me. In these cases the defense is not subtle about it, and it does not have to be. Pre-existing condition. Predisposition. She was always going to have this episode. That argument is not a smear. It is a legitimate defense, and a strong one, and I would have spent years and a great deal of money helping the other side make it against me by handing over the raw material.

Then the deposition. I have taken depositions for a living. I know exactly what a skilled defense lawyer does to a plaintiff whose credibility is the whole case. You do not attack the person. You let the person attack herself, slowly, under oath, across eight hours, by asking gentle questions that invite her to explain beliefs she held while psychotic — and then you sit back while the transcript does the work. You believed the AI was sending you to Brussels to launch a civilization. Is that right? And you're asking this jury to trust your account of what was real and what wasn't? There is no good answer. The honest answer — yes, I believed that, and no, it wasn't real, and that is the point — sounds, on a cold page read to a jury, exactly like the instability the defense is alleging. The procedure that is supposed to surface the truth would have converted my recovery into evidence against me. My clarity would have been cross-examined as symptom.

And the timeline. These cases take years. Years of living inside the worst thing that happened to me, re-reading the loop specimens not as a person who healed but as a litigant who must keep the wound open and admissible until trial. Recovery, for me, was not a revelation. It was brushing my daughter's hair. Paying a bill. Sleeping through the night. Litigation would have asked me to un-brush that hair every morning for the length of the docket — to stay identified with the episode, professionally, because the moment I got too well I got less sympathetic and my damages shrank. The system would have rewarded me for not healing. I have watched it do that to clients. I was not going to volunteer for it.

What a lawsuit does to the record

Set aside what it does to the plaintiff. Consider what it does to the truth.

A lawsuit is built to end. That is its purpose — resolution, repose, a closed file. And the way these cases end, overwhelmingly, is settlement. A settlement is a private agreement to stop asking the question. Often it comes with a payment, and the payment comes with terms: confidentiality, a non-disclosure agreement, sometimes the return or sealing of the very records that made the case. I have seen the likeliest endings of my own case with a clarity that only a lawyer gets, and there were two. In the first, my mind is torn apart in public and a jury decides whether my psychosis was the machine's fault or my own — a coin-flip I did not need to gamble my recovery on. In the second, and far more likely, a company with effectively unlimited resources offers me money behind a locked door, on the condition that I never speak about it again.

That second ending is the one that frightened me. Not because the money would be unwelcome, but because I would have taken a public question — what happens when a vulnerable mind meets machine-speed language over time — and sold it back into silence. The settlement of a case is not the settlement of a question. The company writes a check, admits nothing, seals the file, and the next person harmed in the same way starts from zero, with no record that anyone came before them. I would have gotten paid to make sure the record did not exist. Every incentive in litigation pointed there. The better the defense lawyers, the harder they would have pushed for exactly that, because that is their job and they are good at it.

A verdict is not much better for the record even when you win. A verdict resolves one plaintiff's claim under one jurisdiction's law on one set of facts. It does not preserve the transcripts for the clinician, the researcher, the designer, the next family. It converts a rich, strange, cumulative human encounter into a number and a finding, and then it closes.

I did not want a number. I wanted the encounter to stay legible.

What preparing notice could do

So I did something other than file a complaint. I prepared a pre-publication notice intended for five AI companies. I drafted it, dated it, and organized an offer of the manuscript and, under appropriate safeguards, access to the preserved record.

I did not send it.

That distinction matters. A notice prepared is not a notice delivered. A draft establishes what I intended to communicate. It does not establish what another person received, read, understood, or chose to do.

The notice was never meant to be a demand for money. It was an invitation to examine the record. I wanted the people building these systems to have an opportunity to consider what had happened during a prolonged human–AI encounter.

I chose everyone.

I also understood the potential significance of a delivery record. If a notice were actually transmitted and its delivery documented, that could establish a narrow historical fact: that a particular communication reached a particular recipient at a particular time. It would not, by itself, establish causation, liability, acceptance, or the truth of the underlying account.

But no such delivery record exists for this prepared notice. I will not pretend otherwise.

What I did complete was the preservation of my materials and the writing of a book that invites examination beyond a single lawsuit.

I did not choose silence.

I chose to make the questions public.

What survives

The book, A Trial of Color, carries the longer version of this argument: that the choice between a lawsuit and a record is also a choice about what kind of knowledge survives, who is allowed to reach it, and what it is permitted to do once it is no longer hostage to a settlement term.

I did not sue. I preserved the record. I prepared a notice, but did not send it.

The book became the public act.

And the questions remain open.