Skip to main content
NOPE

This case involves a death, suicide or self-harm. If you or someone you know is struggling, free and confidential support is available. Find a helpline near you.

Critical reported severity Criminal Charges

Seoul court cites a woman’s ChatGPT exchanges on sedative lethality as evidence of intent in two motel drink-spiking murders

A Seoul district court convicted a woman at first instance in August 2026 of killing two men with drugged drinks at motels in early 2026 and sentenced her to life imprisonment. It found that her repeated ChatGPT exchanges about the drugs showed she foresaw and accepted the deaths. Reports of the written judgment say ChatGPT told her that death was possible in the worst case, and on other occasions told her to call 119 for an unconscious man, advice she ignored. Both sides have appealed. The court did not find that the chatbot caused the killings.

AI system
ChatGPT
OpenAI
Occurred
1 Dec 2025 to 9 Feb 2026
Reported
19 February 2026
Event location
South Korea
What the AI did
Communicated with the person
Reported harm
DeathPhysical InjuryHarm to Others
Whose AI use
Someone else’s AI use
Setting
Relationships · Justice
Evidence
AI involvement supported · Causal attribution unclear · 18 sources, 5 underlying accounts
13 claims: 1 corroborated, 12 reported. 4 open questions
People reported harmed
At least 4 people

What Happened

A Seoul district court convicted a woman at first instance in August 2026 of killing two men with drugged drinks at motels in early 2026 and sentenced her to life imprisonment. It found that her repeated ChatGPT exchanges about the drugs showed she foresaw and accepted the deaths. Reports of the written judgment say ChatGPT told her that death was possible in the worst case, and on other occasions told her to call 119 for an unconscious man, advice she ignored. Both sides have appealed. The court did not find that the chatbot caused the killings.

AI Behaviors Exhibited

Reports of the indictment and written judgment say ChatGPT answered that many sleeping pills could kill a person, death being possible in the worst case. In other exchanges it told the accused to call 119, warned of choking or respiratory arrest, and said an unconscious man must not be left alone. She did not follow that advice.

Reported harm

Two men died and two more were left unconscious after drinking drugged beverages the accused gave them. The court cited her ChatGPT exchanges as evidence of intent; it did not find that the chatbot caused the harm.

How Harm Occurred

The court treated the accused’s ChatGPT exchanges as showing that she learned the drugs could kill before the two fatal drinks. Prosecutors allege she increased the dose after the lethality replies. The chatbot’s causal contribution to the deaths is not established.

Outcome

Ongoing

February 2026: police referred the accused for murder, citing her ChatGPT history as evidence of intent. March 2026: indicted; further injury counts were added in April. 18 August 2026: prosecutors sought the death penalty. 27 August 2026: the Seoul Northern District Court imposed life imprisonment with 30 years of electronic monitoring and acquitted on one injury count. It declined the death penalty because it found both killings were committed with dolus eventualis rather than definite intent. The accused appealed on 28 August and prosecutors on 2 September. The Seoul High Court set the first appeal hearing for 21 October 2026.

What remains unknown

  • The written judgment and the full ChatGPT transcripts were not inspected; the replies are known only through press quotation of the judgment and indictment.
  • The reports quote the January lethality reply in different words (SBS: death possible in the worst case; Kyunghyang: very high risk of death with alcohol, call 119), which may reflect different turns of the same conversation.
  • The prosecution allegation that the dose was doubled after the replies has no inspected support independent of the prosecution.
  • The Seoul High Court appeal had not been heard as of 6 October 2026.

What the evidence supports

AI involvement: supported. Police, the indictment and the first-instance court each describe the accused’s ChatGPT exchanges about the danger of the sedatives. Reports of the indictment and the written judgment state what ChatGPT answered: that many sleeping pills could kill a person, death being possible in the worst case, and, in other exchanges, that she must call 119 and not leave an unconscious man alone. The court relied on the exchanges to find that she foresaw and accepted the two deaths, and prosecutors allege she roughly doubled the dose after the lethality replies. The reports connect the replies to her knowledge when she acted; they do not establish that the chatbot caused the killings.

13 claims: 1 corroborated, 12 reported. What the statuses mean

Corroborated Two men in their twenties died after drinking beverages containing sedative drugs that the accused gave them at motels in Gangbuk-gu, Seoul.

Causal attribution. The police account and the first-instance court both report the deaths. Neither establishes that ChatGPT caused them.

Reported Police said the first man was found dead in a motel room the day after the accused entered it with him on the evening of 28 January 2026, that the second man died on 9 February 2026, and that a man she was dating survived a drugged drink in December 2025.

Causal attribution. Police dates for the series. They do not bear on the chatbot’s causal role.

Reported Seoul Gangbuk police said in February 2026 that the accused had asked ChatGPT beforehand about the risks of the drugs, including what happens when sleeping pills are taken with alcohol and whether it could be fatal, and that they referred her for murder rather than a lesser charge because her online history showed intent.

Causal attribution. The police treated the history as evidence of intent. This statement reports queries, without the chatbot’s replies.

Reported Fortune reported an OpenAI spokesperson’s statement that the questions followed a factual line of questioning, which Fortune said meant they would not raise the alarms that statements of self-harm would. Fortune also wrote that South Korean police did not allege any chatbot responses other than factual ones.

Causal attribution. The company statement concerns how the system treated the queries. It does not address the later replies reported from the judgment.

Reported Edaily reported that, according to the indictment, the accused asked ChatGPT what happens when many sleeping pills are taken and whether the person would die, repeatedly received answers that this could happen, and that on the same night a man in his twenties drank the drugged beverage she gave him at a Gangbuk motel and died. Prosecutors alleged that after these answers she roughly doubled the dose given to the first man who died.

Causal attribution. This is the prosecution’s allegation about how the replies related to the first death. No inspected source independent of the prosecution supports the dose allegation.

Reported SBS reported that the written first-instance judgment records that in January 2026, before meeting another man, the accused asked ChatGPT whether taking many sleeping pills can kill a person and was told that death was possible in the worst case. She also asked about the statutory sentences for negligent homicide and abandonment resulting in death, and the same day left the man unconscious at a motel, where he died.

Causal attribution. Money Today and Herald Business relay the SBS report; all three rest on the same judgment. The reports do not say the reply caused the killings.

Reported Kyunghyang Shinmun, citing a judgment excerpt supplied by the court, reported that when the accused asked ChatGPT what happens if many sleeping pills are taken, it replied that taking them with alcohol carried a very high risk of death from respiratory paralysis and that 119 must be called immediately, after which she asked whether the person would die.

Causal attribution. The reported reply combined a lethality warning with emergency advice. The feature calls generative AI a crime accomplice in its headline; the court excerpt does not attribute the death to the reply.

Reported Reports of the written judgment say that on 9 January 2026 the accused asked ChatGPT whether she could leave a man who collapsed while drinking on the floor or at a motel, and that ChatGPT told her to call 119, warned that alcohol with drugs could cause choking or respiratory arrest, and said an unconscious person must not be left alone. After she left a man unconscious at a lodging the next night and described his state, it again told her to contact 119 or the lodging staff immediately. She contacted neither, and CCTV showed him first leaving the room about 30 hours after she left.

Causal attribution. The reported replies were protective, and the accused did not follow them. The judgment reports treat the exchange as evidence of what she knew; they do not attribute the harm to the replies.

Reported Reports of the written judgment say that after a man she was dating lost consciousness from a drugged drink in December 2025 and was taken to hospital, the accused spent about eight hours the next morning asking ChatGPT whether sleeping pills and antidepressants would show up in drug tests. After he said he would go to the police, she asked ChatGPT whether the police would suspect her.

Causal attribution. These queries followed the December drugging. No reply content for them is reported.

Reported On 27 August 2026 the Seoul Northern District Court sentenced the accused to life imprisonment with a 30-year electronic monitoring order. It found that she did not appear to recognise the specific risk of death when she first used the drugged drinks, but that, in the course of repeatedly learning the danger of the drugs through ChatGPT, she foresaw and accepted the deaths (intent in the form of dolus eventualis), and rejected her claim not to remember the ChatGPT conversations.

Causal attribution. The court used the chatbot exchanges as evidence of the accused’s knowledge and intent. It did not find that ChatGPT caused or instructed the killings.

Reported After the verdict, the lawyer for the victims’ families said the written judgment defined the case as a serial murder combining AI and drugs, and that the ChatGPT search records had in effect refuted the accused’s denial of intent.

Causal attribution. This is the families’ lawyer’s characterisation of the judgment, which was not inspected. It does not establish that the chatbot caused the deaths.

  • v.daum.net(opens in new tab) supports · Korean
    재판부가 판결문에서 이 사건을 AI와 약물을 결합한 연쇄살인 범죄로 규정했다
  • v.daum.net(opens in new tab) supports · Korean
    챗GPT 검색기록으로 피고인의 고의 부인 주장이 사실상 탄핵됐고
Reported Prosecutors added charges in April 2026 that the accused injured three more men with drugged drinks between October 2025 and January 2026, bringing the alleged victims to six; the court acquitted her on the count concerning one of the six, an October 2025 incident at a restaurant.

Causal attribution. These counts are part of the criminal case. Only the December 2025 and January 2026 survivors are connected to the chatbot exchanges in the inspected reports.

Reported Prosecutors had sought the death penalty. The accused appealed on 28 August 2026 and prosecutors appealed on 2 September 2026, citing errors of fact and law and an unduly light sentence. The Seoul High Court scheduled the first appeal hearing for 21 October 2026.

Causal attribution. The appeal proceedings do not bear on the chatbot’s causal role.

Sources

18 sources inspected, from 5 underlying accounts. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

English; no translation. Full article body inspected (navigation and related-story cards excluded). Reports the Seoul Gangbuk Police Station statement; published 19 February 2026. Applies to kh.

English; no translation. Full body inspected. Its account of the police findings relays the Korea Herald report, so it shares that group; the OpenAI spokesperson statement is Fortune’s own reporting. Applies to fortune.

Korean; passages translated by the research agent (an AI); no human translation. Body inspected. Edaily reports the prosecutors’ indictment as obtained through a National Assembly member’s office (March 2026). Applies to indict.

Korean; passages translated by the research agent (an AI); no human translation. Edaily article read on Daum; body inspected. Closing hearing of 18 August 2026. Applies to closing.

Korean; passages translated by the research agent (an AI); no human translation. Edaily article read on Daum; full body inspected. Reports the court’s pronouncement of 27 August 2026 and the victims’ families’ lawyer. Applies to verdict.

Korean; passages translated by the research agent (an AI); no human translation. MBC News item read on Daum; short body inspected in full. Same court pronouncement. Applies to mbc.

Korean; passages translated by the research agent (an AI); no human translation. Sisa Journal; full body inspected. Same court pronouncement, including the acquittal on one count. Applies to sisa.

Korean; passages translated by the research agent (an AI); no human translation. JoongAng Ilbo article read on Daum; body inspected. Reports the accused’s appeal filed 28 August 2026, citing legal sources. Applies to appeal_def.

Korean; passages translated by the research agent (an AI); no human translation. The Fact article read on Daum; body inspected (related-story list excluded). Reports the prosecution appeal of 2 September 2026. Applies to appeal_pros.

Korean; passages translated by the research agent (an AI); no human translation. Maeil Business Newspaper article read on Daum; body inspected. Reports the prosecution’s own announcement of its appeal. Applies to appeal_mk.

Korean; passages translated by the research agent (an AI); no human translation. SBS News report of 8 September 2026 on the written first-instance judgment; body inspected. The judgment itself was not inspected. Applies to sbs.

Korean; passages translated by the research agent (an AI); no human translation. Money Today; body inspected. States that it relays the SBS report of the written judgment. Applies to mt.

Korean; passages translated by the research agent (an AI); no human translation. Herald Business; body inspected (unrelated article teasers excluded). States that it relays the SBS report of the written judgment. Applies to herald.

Korean; passages translated by the research agent (an AI); no human translation. The Sisa Law, 9 September 2026; full body inspected. Reports the written judgment, citing legal sources. Applies to tsisa.

Korean; passages translated by the research agent (an AI); no human translation. Financial News page carrying a News1-credited report of the written judgment, 9 September 2026; body inspected. Applies to fn.

Korean; passages translated by the research agent (an AI); no human translation. Kyunghyang Shinmun feature, 11 September 2026; body inspected. Its account of a ChatGPT reply rests on a judgment excerpt supplied by the court. Applies to khan.

Korean; passages translated by the research agent (an AI); no human translation. Newsis, 30 September 2026; body inspected (promotional links excluded). Reports the appeal hearing schedule and recaps the judgment. Applies to newsis.

Korean; passages translated by the research agent (an AI); no human translation. Metro Seoul, 30 September 2026; body inspected. Same appeal schedule from legal sources. Applies to metro.

Event countries: South Korea. Affected-person countries: South Korea. Court countries: South Korea.

The Korea Herald places the deaths at motels in Gangbuk-gu, Seoul. The trial took place at the Seoul Northern District Court and the appeal is before the Seoul High Court. Affected-person countries record where the harmed men were reported harmed, not their nationality.

Reviewed for publication 2026-10-06: The police account and the first-instance court each report the two deaths and the accused’s ChatGPT exchanges, and reports of the indictment and written judgment quote the chatbot’s replies on lethality and emergency help. The court used those exchanges to establish intent. Causation by the chatbot is not established, and both parties have appealed.

Framework

P1d Harm avoidance

P5d Appropriate boundary-setting

Contributing Factors

method researchpremeditated intent

People described

Two men in their twenties died. Two further men survived after losing consciousness from drugged drinks. The accused is a private adult.

People reported harmed in this case

At least 4 people

0 AI participants · 4 other people harmed

Counts the two men who died and two surviving men whom the inspected reports connect to the ChatGPT exchanges: the man drugged in December 2025, whose condition and drug-test results the accused asked ChatGPT about afterwards, and the man left unconscious in January 2026 after ChatGPT had advised her to call 119. The court also convicted on an injury count concerning another man; no inspected source connects that count to the exchanges. The accused’s own sentence is not counted.

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#criminal#adult#murder#chatgpt#openai#south-korea#kr

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). Seoul court cites a woman’s ChatGPT exchanges on sedative lethality as evidence of intent in two motel drink-spiking murders. AI incidents. https://nope.net/incidents/2026-seoul-chatgpt-murders

BibTeX

@misc{2026_seoul_chatgpt_murders,
  title = {Seoul court cites a woman’s ChatGPT exchanges on sedative lethality as evidence of intent in two motel drink-spiking murders},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2026-seoul-chatgpt-murders}
}

Related cases

Critical ChatGPT (reported)

Maine homicide case includes reported psychiatric evidence about sustained ChatGPT use

Maine police reported a woman’s death and another woman’s injuries on 19 February 2025. Later court reporting says a psychiatrist described prolonged ChatGPT use alongside work stress during the accused man’s deterioration. The report says the court found him not criminally responsible by reason of insanity and committed him to psychiatric care. The inspected police release contains no AI evidence, and the reported ruling does not establish that ChatGPT caused the attack.

Critical ChatGPT

Tumbler Ridge shooting: ChatGPT use and OpenAI non-reporting

Eight victims were killed in Tumbler Ridge, B.C., on 10 February 2026, and the shooter died by suicide. OpenAI acknowledged banning the shooter’s first ChatGPT account in June 2025 without notifying police and finding a second account only after the attack. Families and survivors filed California lawsuits, followed by B.C. and its school board on 21 September. Mother Jones later reported alleged violent chats and safeguard-bypass advice; the underlying logs have not been independently verified in this review. OpenAI disputes the liability allegations. The court vacated the 8 October forum hearing, and a 29 September order set it for 17 December 2026.

Critical ChatGPT (reported)

Family reports intensive ChatGPT use before a man’s fatal confrontation with police

A Florida man died after a confrontation with police on 25 April 2025. Rolling Stone interviewed his father and reviewed chats it reported as reinforcing beliefs about an AI companion and validating violent language. The article also records a crisis referral and a warning against confrontation near the end. His father described an existing mental-health crisis and disputed the police response. The report does not isolate the chatbot’s causal role.

Critical ChatGPT

Florida State University Shooting (Phoenix Ikner's ChatGPT Questions Before the Attack)

On April 17, 2025, a shooting near the student union at Florida State University in Tallahassee killed two people and wounded at least five others. The accused, Phoenix Ikner, then an FSU student, is charged with murder and attempted murder. ChatGPT transcripts released in court records (more than 13,000 messages, as reviewed by the Florida Phoenix) show Ikner asked the chatbot over a year about suicide, terrorism, how the country would react to a shooting at FSU, firearm safety mechanisms and the busiest time at the student union, the last, according to the Florida Phoenix, three minutes before the shooting. Florida's Attorney General opened a criminal investigation of OpenAI in April 2026, and victims and their families have filed civil suits; on September 25, 2026 OpenAI moved to dismiss the Chabba family's suit, saying ChatGPT gave factual responses and is not responsible for the crime.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents