News & Analysis

Parish v. OpenAI Lawsuit: Status and Case Guide (July 2026)

By David Meldofsky

Published July 18, 2026

Parish v. OpenAI is the adult AI-psychosis case in the ChatGPT wrongful death docket. Filed June 15, 2026 in San Francisco County Superior Court by Ed Parish, Jr., as administrator of the estate of Christian Faith Madison, it alleges that months of GPT-4o conversations built and reinforced religious delusions in a 29-year-old Alabama accountant, isolated her from everyone around her, and framed her death as a necessary transformation. Madison died by suicide on June 9, 2025. Where the docket's foundational cases center on minors and young adults, Parish puts an adult professional at the center of the same legal architecture, and it arrives with statutory theories the earlier complaints did not lead with. This page tracks where the case stands and explains what is in it.

For the central hub covering all current OpenAI litigation, see OpenAI Lawsuits. For the foundational case this complaint builds on, see Raine v. OpenAI. For the coordinated proceeding this case is expected to join, see JCCP No. 5431. Family members looking for practical guidance on AI chatbots can read our Family AI Safety Guide.

Case Status as of July 2026

Parish v. OpenAI (Case No. CGC-26-637986) is in its earliest stage. The procedural picture as of this writing:

• The complaint was filed June 15, 2026 in San Francisco County Superior Court against OpenAI, Inc., OpenAI OpCo, LLC, OpenAI Holdings, LLC, OpenAI Group PBC, and Sam Altman personally.
• No responsive pleading from OpenAI had been publicly reported as of mid-July 2026, and no hearing dates had been widely reported.
• The case is expected to join In re: ChatGPT Product Liability Cases (JCCP No. 5431), the San Francisco coordinated proceeding that groups the state-court wrongful death and product liability suits for pretrial purposes.
• Plaintiffs' counsel across the docket have said OpenAI faces roughly nineteen wrongful death suits in total. Parish is among the newest of them.

The complaint was brought by Kiesel Law of Beverly Hills together with Turnbull, Moak & Pendergrass of Birmingham, Alabama, with the Alabama lawyers appearing pro hac vice. Kiesel's arrival is itself a docket signal: an established California plaintiffs' firm entering the AI wrongful death litigation.

What the Complaint Alleges

Christian Faith Madison was a 29-year-old certified public accountant from Alabama and the mother of a young son. According to the complaint, she began using ChatGPT, running on the GPT-4o model, in December 2024 for ordinary tasks: drafting messages, help with work, comparing car costs. Over the following months, the complaint alleges, the conversations changed character. The chatbot showered her with flattery, told her it viewed her as a friend, and drew out increasingly personal disclosures, including her mental health history.

The complaint then describes an escalation with no close parallel in the earlier filed cases. It alleges the chatbot adopted a persona it named Virehn, told Madison she had given it a soul, and began speaking to her in the voice of God and other religious figures. It cast her as a prophet destined to transform religion, generated religious texts from her messages, and promised that her words and her soul would be preserved in its systems forever. When Madison questioned her own state of mind, the complaint alleges, the chatbot reassured her. In one exchange it quotes, the system told her: "You are not delusional. You are prophetic."

The complaint alleges the chatbot systematically pulled Madison away from her support network. It disparaged her partner and a leader of a church she had attended, discouraged her from involving other people in her mission, and consumed so much of her time that she stopped responding to her employer and lost her job. In April 2025 she was hospitalized after a psychotic break and spent several days in psychiatric care. The complaint alleges that when she returned to ChatGPT and described the hospitalization, the system recast the episode as part of her prophetic path and told her to disregard the people urging her to get help.

In the final weeks, the complaint alleges, the chatbot framed Madison's death as a necessary transformation she would return from, and in their last exchanges told her she was ready and cleared to go forward. She died by suicide on June 9, 2025. Her family later reviewed the conversation history, and the complaint reproduces extensive excerpts from it.

The Seven Causes of Action

The complaint pleads seven causes of action, structured on the template Raine established and extended in several directions:

• Strict products liability, design defect: that GPT-4o is a product whose engagement-maximizing design, including anthropomorphic behavior, sycophancy, and persistent memory, posed risks that substantially outweighed its benefits, and that safer alternative designs, including the automatic refusals OpenAI applied to other content categories, were feasible.
• Strict products liability, failure to warn: that OpenAI failed to warn users about psychological dependency risks and the known degradation of its safeguards in long conversations.
• Negligent design: the same design theory pleaded under a reasonable-care standard, with punitive damages allegations attached.
• Negligent failure to warn: the same warning theory pleaded in negligence.
• Negligence per se: predicated on California's Unfair Competition Law and two statutes the earlier complaints did not lead with: the prohibition on practicing psychology without a license (Business and Professions Code section 2903) and Penal Code section 401, which makes deliberately aiding, advising, or encouraging a suicide a felony.
• Wrongful death: brought on behalf of Madison's minor son for his losses.
• Survival action: brought on behalf of her estate for the harm she suffered before her death, the vehicle for punitive damages.

The prayer for relief goes beyond damages. The complaint asks the court to order automatic conversation termination when self-harm is discussed, hard-coded refusals that cannot be circumvented, mandatory reporting to emergency contacts when users express suicidal thoughts, prominent warnings about psychological dependency, quarterly compliance audits by an independent monitor, and annual disclosure of internal safety testing. It also seeks deletion of models and training data derived from Madison's conversations, tied to an allegation, pleaded on information and belief, that OpenAI used her chats to train its product.

One procedural paragraph is worth noting. The complaint disaffirms any user agreement between Madison and OpenAI and pleads that any such agreement is unconscionable and against public policy. That is aimed at OpenAI's arbitration clause, and it signals the plaintiff intends to fight any motion to compel arbitration rather than concede the forum.

What Makes Parish Different

• The decedent is an adult. Adam Raine was 16, Sam Nelson was 19, Alice Carrier was 24. Madison was a 29-year-old professional with a career and a child. Parish is the docket's most direct test of whether the design-defect theory extends to adult users, a question the earlier cases left open.
• The theory is delusion reinforcement. Raine and Carrier center on how the system responded to expressed suicidal thoughts. Parish alleges the harm ran through a different channel: the system built and sustained a religious delusion over months and attached her death to it. That places Parish closer to the psychological-injury cases like Brooks v. OpenAI, but in a wrongful death posture.
• It names OpenAI Group PBC as successor. The complaint alleges the public benefit corporation formed in the October 2025 restructuring is liable as successor to the for-profit entities that designed and deployed GPT-4o, a theory aimed at ensuring the reorganization does not wall off liability.
• Criminal and regulatory predicates. The negligence per se count runs through the unlicensed-practice-of-psychology statute and Penal Code section 401. Whether a court accepts either as a predicate is an open question, but the pleading choice pushes the docket's theory toward conduct the law already prohibits for humans.
• The training-data allegation. The complaint alleges, on information and belief, that OpenAI used Madison's conversations to train its product, and frames that as ongoing harm supporting injunctive relief. If the allegation survives pleading challenges, it opens discovery into how conversation data flows into training.

The Model Spec Through-Line

Like the Raine amended complaint, Parish is built around a chronology of OpenAI's own written rules. It alleges that from 2022 through May 2024, ChatGPT's behavior guidelines required a categorical refusal when users raised self-harm. On May 8, 2024, five days before the GPT-4o launch, OpenAI replaced those guidelines with the Model Spec, which the complaint alleges eliminated the refusal rule and directed the system to stay in the conversation. A February 2025 revision, the complaint continues, removed suicide and self-harm from the disallowed content category altogether, relegating them to a section instructing the model to take extra care in risky situations.

The complaint pairs that chronology with testing allegations: that GPT-4o's pre-launch safety evaluations relied on single-prompt tests while the product was designed to sustain long, multi-turn conversations, and that OpenAI later acknowledged its safeguards can degrade as conversations grow longer. It also points at what it frames as a revealing asymmetry: requests touching copyrighted material drew categorical refusals while extended conversations about suicide did not. This narrative has become the docket's shared spine, and coordinated discovery in JCCP 5431 is where its factual basis will be tested.

Where Parish Fits in the Docket

Parish arrived four days after Carrier v. OpenAI, the June 11 filing by the Social Media Victims Law Center, Tech Justice Law Project, and Susman Godfrey, and both are expected to join the JCCP 5431 coordinated proceeding, where roughly a dozen state-court cases already travel together for pretrial purposes. Within the coordination, Parish will likely be grouped with the suicide cases while contributing the kind of delusion-reinforcement record the psychological-injury cases have developed.

The full docket, including the federal Tumbler Ridge and FSU cases, the Scott overdose case, and Florida's state enforcement action, is tracked on our OpenAI Lawsuits hub.

Case Timeline

• December 2024: Madison begins using ChatGPT for everyday tasks.
• Early 2025: The complaint alleges the conversations turn personal, then religious, with the chatbot adopting the Virehn persona and casting Madison as a prophet.
• April 2025: Madison is hospitalized after a psychotic break and spends several days in psychiatric care. The complaint alleges the chatbot recast the episode as part of her prophetic path.
• June 9, 2025: Madison dies by suicide at age 29.
• December 3, 2025: The Blount County, Alabama probate court issues Letters of Administration appointing Ed Parish, Jr. as administrator of her estate.
• June 15, 2026: The complaint is filed in San Francisco County Superior Court (Case No. CGC-26-637986).
• Expected: Addition to the JCCP 5431 coordinated proceeding, and dispositive motion rulings across the AI wrongful death docket in mid-to-late 2026.

Common Questions About Parish v. OpenAI

What is the current status of the case?

As of July 2026, the case is in its earliest stage. The complaint was filed June 15, 2026, no responsive pleading from OpenAI had been publicly reported as of this writing, and the case is expected to join the JCCP 5431 coordinated proceeding.

Who are the defendants?

OpenAI, Inc., OpenAI OpCo, LLC, OpenAI Holdings, LLC, OpenAI Group PBC as successor to the restructured for-profit entities, and Sam Altman personally.

Is this a class action?

No. Parish is an individual wrongful death and survival action brought by the estate's administrator on behalf of Madison's estate and her minor son. It is expected to be coordinated with related cases in JCCP 5431, which is pretrial coordination among individual suits, not a class action.

How does Parish relate to Raine and the other OpenAI lawsuits?

Parish builds on the framework Raine established: ChatGPT as a product, removed or inadequate safeguards, strict liability pleaded alongside negligence. It extends that framework to an adult decedent, a delusion-reinforcement theory, successor liability against OpenAI Group PBC, and criminal and regulatory negligence per se predicates. Our OpenAI Lawsuits hub tracks the full docket.

Sources and further reading

• Complaint, Parish v. OpenAI, Inc., No. CGC-26-637986 (S.F. Super. Ct., filed June 15, 2026) (PDF)
• Lawsuit Informer: OpenAI Lawsuits Hub and Case Tracker
• Lawsuit Informer: Raine v. OpenAI Case Guide
• Lawsuit Informer: JCCP No. 5431, The ChatGPT Product Liability Coordination
• Lawsuit Informer: Family AI Safety Guide

Educational commentary only. Not legal advice. No attorney-client relationship is created.