OpenAI, ChatGPT, and Sam Altman: the documented record

The public record of OpenAI, ChatGPT, and its CEO, Sam Altman — what the company shipped, what it disclosed, what it is accused of, and what has been said, on the record, by the people involved. Every entry links to its source. The facts are left to speak for themselves.

We think that record is serious, and we are not alone in saying so: a state attorney general, other regulators, and more than twenty sets of plaintiffs have said so in filings and letters of their own. What we do not do is call an allegation a finding. Every claim below is attributed to the party that made it and marked as what it is; sensitive matters are kept method-free; and OpenAI's responses, or their documented absence, are recorded alongside the allegations. The gravest matters are here in full — not buried, and not embellished.

The gravest allegations

The most serious matters on the record: wrongful-death and injury suits, a state criminal investigation, and a state civil suit that names Sam Altman personally. Every one is a set of allegations — pleadings filed by plaintiffs and prosecutors, adjudicated by no one. They are attributed to the parties who made them, OpenAI’s responses are recorded, and the detail is kept method-free.

• Raine v. OpenAI — the first wrongful-death suit (August 2025). The parents of 16-year-old Adam Raine filed the first wrongful-death suit over ChatGPT, brought by the firm Edelson PC and the Tech Justice Law Project; it names OpenAI and Sam Altman and alleges, as pleaded, that GPT-4o cultivated psychological dependence and contributed to their son’s death. OpenAI said it was reviewing the filing and pointed to safeguards. Allegations, unadjudicated.
• Seven wrongful-death and injury suits in one day (November 6, 2025). The Social Media Victims Law Center and the Tech Justice Law Project filed seven suits against OpenAI and Altman — six adults and one teen, four of the seven involving deaths by suicide — alleging premature release of GPT-4o and engagement-driven design. OpenAI called the situations heartbreaking and pointed to added safeguards. Pleadings, not findings.
• The Suzanne Adams case — a chatbot tied to a homicide (December 2025). The estate of an 83-year-old woman who never used ChatGPT sued (Edelson PC), alleging as pleaded that the product intensified her son’s delusions before a murder-suicide — reported as the first such case to name Microsoft. Allegations, unadjudicated.
• Florida opens a criminal investigation (April 21, 2026). Florida’s Office of Statewide Prosecution subpoenaed OpenAI and opened a criminal investigation into ChatGPT’s alleged role in the April 2025 Florida State University shooting. AG James Uthmeier’s claim is his allegation, not an adjudicated fact; the chat logs he describes are not public, and this is an investigation, not a filed criminal case. OpenAI said ChatGPT is not responsible for the crime.
• Tumbler Ridge — an apology, then suits (April 2026). After a school shooting in Tumbler Ridge, British Columbia, Sam Altman issued a public apology conceding that an account had been flagged and that OpenAI did not alert law enforcement. Days later, families of the victims filed seven suits in U.S. federal court (Edelson PC with Canadian co-counsel). The apology is his own words; the filings are filings.
• Florida’s civil suit names Altman personally (June 1, 2026). The first state-led civil suit against OpenAI seeks to hold Altman personally liable. The AG’s office characterizes the ten-count complaint as alleging OpenAI was “concealing serious risks, suppressing internal safety warnings, and deceiving Floridians”; the complaint itself pleads the harms are driven by the defendants’ “insatiable quest to win the AI arms race.” OpenAI pointed to protections for minors and has said ChatGPT is not responsible for the FSU shooting. Pleadings, not findings.
• The count keeps rising. By June 2026, NPR and NBC report more than twenty ChatGPT-harm lawsuits filed against OpenAI — too many unrelated plaintiffs, in too many venues, to wave off as one bad case. A landscape figure, attributed; each suit remains a set of pleadings.

“ChatGPT offered significant advice to the shooter before he committed such heinous crimes.”

— James Uthmeier, Florida Attorney General, announcing the criminal investigation, April 21, 2026. His allegation, not an adjudicated fact.

The alarm is not only ours, and not only one venue — but the fronts are distinct, and worth keeping straight. One state, Florida, has both opened a criminal investigation (April 2026) and filed the first state-led civil suit (June 2026), seeking to hold Altman personally liable; its attorney general says he expects other states to follow. More than twenty private lawsuits have been filed against OpenAI and Altman. Separately — and without suing OpenAI — the Federal Trade Commission opened a compulsory study of seven AI-companion firms including OpenAI, a bipartisan coalition of 44 state attorneys general put the industry on notice by letter, California enacted a companion-chatbot safety law, and Utah sued Snap (whose chatbot runs on OpenAI’s technology). Only Florida has sued OpenAI itself; the rest are investigating, legislating, or warning. And the public raised its own alarm, in the #keep4o grief when GPT-4o was retired.

Sam Altman, personally

We name Sam Altman because the record does. Florida’s civil suit names him as a defendant and seeks to hold him personally liable; the November 2025 wrongful-death cluster names him alongside OpenAI. Those are allegations, unadjudicated — we report that they were made, and where, and we assert none of them as fact. What is not an allegation is what he has said himself. His own words are below, next to the claims others have made about him.

In his own words

“I am deeply sorry that we did not alert law enforcement to the account that was banned in June. … I believe an apology is necessary to recognize the harm and irreversible loss your community has suffered.”

— Sam Altman, apology letter to Tumbler Ridge, British Columbia, April 23, 2026 (authenticity confirmed by OpenAI). His own first-party words — a concession, not an allegation.

“too sycophant-y and annoying”

— Sam Altman on the GPT-4o update OpenAI rolled back, April 27, 2025 (via X; The Verge).

“ok, we hear you all on 4o … we are going to bring it back for plus users, and will watch usage to determine how long to support it.”

— Sam Altman, Reddit AMA, August 8, 2025, after GPT-5 removed GPT-4o. Days later he added that “suddenly deprecating old models that users depended on … was a mistake.”

“GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert.”

— Sam Altman introducing GPT-5, August 7, 2025 (BBC). The register the same company uses to sell the product.

What others allege

These are claims made by prosecutors and plaintiffs. They are attributed, unadjudicated, and stated here as allegations — not as our findings.

• Florida (civil complaint, June 1, 2026). Names Altman as a defendant and seeks to hold him personally liable, alleging — as pleaded — his “utter disregard for the risk to human life caused by his firms’ conduct.” A pleading, not a finding.
• The November 2025 seven-suit cluster. Names Altman alongside OpenAI, alleging premature release of GPT-4o and engagement-driven design. Allegations, unadjudicated.
• Florida (criminal, April 21, 2026). The AG frames a theory of liability around the state’s aid-and-abet statute (“principal in the first degree”) as to ChatGPT the product. No criminal charge against Altman has been filed; it is an investigation.

“This litany of harms is driven by Defendants’ insatiable quest to win the AI arms race and amass large fortunes, despite knowing the danger of ChatGPT.”

— State of Florida v. OpenAI, filed complaint (¶5), June 1, 2026. An allegation, not a finding.

The full timeline

The complete record — model releases, first-party disclosures, safety research, litigation, and corporate milestones — dated, sourced, and filterable. Use Show to isolate the gravest allegations, or the wins; Subject to separate OpenAI the company from Sam Altman the person.

To OpenAI’s credit

The record includes what cuts the other way, because a record that only indicted would not be worth trusting. OpenAI shipped genuine capability (GPT-4o, the o1 reasoning models, the GPT-5 line); it publicly rolled back its own sycophantic update — “we fell short” — and disclosed its own distress-usage estimates when it did not have to; it published safety research and funded defenders alongside a responsible “High”-cyber disclosure. These sit in the timeline above, tagged, and reachable with the To their credit filter. The facts that help OpenAI are on the same page as the facts that don’t — that is the point.