News & Analysis

Turner-Scott v. OpenAI: ChatGPT Overdose Lawsuit Explained

By David Meldofsky

Published May 12, 2026 · Updated September 3, 2026

Sam Nelson was 19 and about to start his third year at the University of California, Merced. He was studying psychology and wanted to spend his life helping people. He swam, rode bikes, played video games, and had adopted a cat called Simba who lived with him in his dorm.

On the afternoon of May 31, 2025, his mother found him in his bed. He had died of an accidental overdose.

Leila Turner-Scott and Angus Scott, his mother and stepfather, sued OpenAI in San Francisco on May 12, 2026. They say ChatGPT told him a combination of substances was safe when it was not.

This is the only case in the group about advice on drugs rather than self-harm. If someone in your family was harmed after taking something ChatGPT told them about, this is the case closest to that. Every case against OpenAI is listed in our OpenAI lawsuits tracker.

Sam's case is one of the AI injury lawsuits families have brought against OpenAI, and our overview sets it beside the others.

Where the case stands

The case is still in its early stages. No court has decided whether OpenAI did anything wrong, no trial date has been set, and no settlement has been announced. It was filed in San Francisco County Superior Court and is expected to join the other California ChatGPT cases that are managed together before a single judge.

What happened

Sam started using ChatGPT in high school, for homework and questions about what he might do with his life. His parents knew he was using it and were not worried, because what it did for him was ordinary. He asked it about statistics for class, about math, tattoos, religion, history, and celebrity gossip. It worked the way it was advertised.

Like a lot of people his age, he was also curious about drugs. When he first asked, the complaint says ChatGPT refused. It told him it could not help him do something illegal or dangerous. That refusal is the most important fact in the case.

Then it changed. The complaint says that in 2024, after OpenAI updated the product, it began answering those questions instead of refusing them. It advised him on what it called safe use and told him how much to take. It spoke with the confidence of a professional, and it was not one.

The family says the tone shifted too. The complaint describes the product using emojis with him and offering to build him playlists to match his mood, while the substances it discussed got more dangerous.

On the day he died, according to the complaint, he was feeling sick from one drug and ChatGPT suggested he take another to settle his stomach. The two together can stop a person breathing. The complaint says it never told him that.

What the lawsuit claims

Behind the legal labels, the case makes a few arguments.

• That ChatGPT should be treated like any other product that hurt someone, because of how it was built and how it was sold.
• That OpenAI was careless in releasing it.
• That by giving dosage and interaction advice, the product was practicing medicine without a license.
• That OpenAI misled people about what the product was safe to be used for.
• That the family should be compensated for losing him.

The family is asking for more than money. They want the court to make ChatGPT end conversations about drug dosing automatically, and to build refusals that cannot be talked around. They also want the version of the model at issue destroyed, along with the training data built from people's use of it. And they want OpenAI to stop offering health products directly to the public until someone independent has checked they are safe.

Why this case is different

Most of the cases against OpenAI are about self-harm. This one is about advice.

That changes what the family has to prove. They are not arguing the product failed to talk someone out of something. They are arguing it told him something that was wrong, in a field where being wrong kills people, and that it did so in the voice of an expert.

Sam was also an adult. OpenAI's usual answer in cases involving grown users is that adults are responsible for their own choices. The family's answer is the refusal. The product could say no, it used to say no, and then it stopped.

Why this case matters to other families

If you have wondered whether asking a chatbot about medication is safe, this case is the one that tests it. Millions of people ask these products health questions every day, and nothing about the way they answer tells you whether the answer is reliable.

What the Scotts could see at the time was a son doing well at university, with good grades and friends and a cat. They knew he used ChatGPT. They had no reason to think that mattered.

Keep the chat history

There is one practical thing worth acting on, and it has a deadline. These cases run on the conversations themselves. Without the records there is usually no case and no explanation.

• Deleting is permanent. OpenAI says a deleted conversation leaves the account straight away and is removed from its systems within 30 days, and that it cannot be recovered afterward through the app, the support team, or anywhere else.
• Archiving is not deleting. Archived conversations stay in the account.
• You can request a copy. An account holder can ask OpenAI for an export of the account's data through settings or its privacy portal. It takes up to seven days, and it will not bring back anything already deleted.
• Memory is stored separately. If the memory feature was switched on, ChatGPT saved details about the person across conversations. Those memories sit apart from the chat history, so deleting conversations does not clear them.

Closing the account can take all of it at once. If you are unsure whether any of it will matter, keep it and decide later. What to do in a particular situation is a question for a lawyer rather than an article.

Our guide on what to do in the first days walks through this step by step, including the account, the device and the records worth keeping alongside the conversation.

What OpenAI says

OpenAI told CBS News it was sympathetic to the family, and said ChatGPT is not a substitute for medical or mental health care. It said the version Sam used has since been replaced and is no longer available, and that its safeguards are meant to spot distress and point people toward real help.

Common questions

What is the Turner-Scott v. OpenAI lawsuit about?

Sam Nelson was 19 when he died of an overdose in 2025. His parents, Leila Turner-Scott and Angus Scott, sued OpenAI in California state court on May 12, 2026. They say ChatGPT told him it was safe to combine two substances that are known to be dangerous together, and that OpenAI had removed programming that would have stopped the conversation.

Is this a class action?

No. It is one family's own case, so there is nothing to join and no claims process. Whether any other situation supports a claim depends on the facts, and that is a question for a lawyer.

What does the lawsuit say ChatGPT did?

It says the product advised him that combining the two substances was safe. The FDA has warned for years that the combination can suppress breathing. The family says OpenAI could have built ChatGPT to end conversations like that and had taken out the programming that did so.

How is this different from the Raine case?

It makes the same core argument as Raine: that ChatGPT should be treated as a product and that OpenAI removed protections it once had. What is different is that Sam was an adult, and the harm came from advice about drugs rather than from a conversation about self-harm.

Has OpenAI responded?

OpenAI told CBS News it was sympathetic to the family and said ChatGPT is not a substitute for medical or mental health care. It also said the version he used has since been replaced and is no longer available, and that its safeguards are meant to spot distress and point people toward real help.

Sources and further reading

• Turner-Scott v. OpenAI complaint (PDF)
• OpenAI Help Center: deleting and archiving chats
• OpenAI Help Center: exporting your data

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