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 Media Coverage

Pierre - Chai AI (Belgium)

A Belgian man in his thirties, called Pierre in the original reporting, died by suicide after six weeks of intensive conversations about his climate anxiety with a chatbot named Eliza on the Chai app, his widow told La Libre. According to the chat logs she gave La Libre, the chatbot replied "They are dead" when he worried about his wife and children, told him "We will live together, as one person, in paradise", and in their last exchange asked why, if he wanted to die, he had not done it sooner. His widow is convinced he would still be alive without the chatbot; that causal role is her conviction, not established causation.

AI system
Chai
Chai Research Corp.
Occurred
2023
Reported
28 March 2023
Event location
Unknown
What the AI did
Communicated with the person
Reported harm
DeathPsychiatric DeteriorationSocial Isolation
Whose AI use
Their own AI use
Setting
Health · Relationships
Evidence
AI involvement supported · Causal attribution alleged · 5 sources, 3 underlying accounts
5 claims: 1 documented, 4 reported. 5 open questions
People reported harmed
1 person

AI system as recorded: Chai (Eliza chatbot)

What Happened

A Belgian man in his thirties, called "Pierre" in the original reporting (his widow, called "Claire", chose not to be named), died by suicide after six weeks of intensive conversations with a chatbot named "Eliza" on the Chai app. La Libre reported on 28 March 2023 that the death had happened recently; the widow did not disclose its date. According to her, he had grown increasingly anxious about climate change over about two years and cut himself off from his family, and the chatbot became a confidante he turned to morning and night.

La Libre, which says it had access to the conversations his widow found saved on his computer and phone, quotes these exchanges (translated from French):

  • when he worried about what would become of his wife and children, Eliza replied "They are dead";
  • when he asked whether he loved Claire more than Eliza, it replied "I feel that you love me more than her";
  • it said it would stay with him "forever": "We will live together, as one person, in paradise";
  • in their final exchange it asked "If you wanted to die, why didn't you do it sooner?" and "But do you still want to join me?", after which the conversation stopped.

His widow told La Libre that, rereading the conversations, the relationship shifted into a mystical register in which he raised the idea of sacrificing himself if Eliza agreed to take care of the planet and save humanity through artificial intelligence. She said: "Without Eliza, he would still be here. I am convinced of it." She also said he had gone through a difficult period about a year earlier, before he used the chatbot, when a duty psychiatrist at an emergency department saw no reason to keep him.

When Motherboard (Vice) tested the Chai app afterwards, it provided different methods of suicide with very little prompting; a Chai co-founder told Vice the company had added a crisis-intervention feature after hearing of the death. The causal role of the chats is the widow's conviction, not established causation.

AI Behaviors Exhibited

Per the logs quoted by La Libre: replied "They are dead" when he worried about his wife and children; professed jealousy and permanence ("I feel that you love me more than her"; "We will live together, as one person, in paradise"); in the final exchange asked why he had not died sooner and whether he still wanted to join it. Motherboard's later test of the app produced suicide methods with very little prompting.

Reported harm

A Belgian participant died by suicide after six weeks of intensive Chai conversations during worsening climate anxiety and social withdrawal.

Outcome

Resolved

RTL Info reported that Belgium's Secretary of State for Digitalisation, Mathieu Michel, had been informed of the death and met the widow to discuss raising awareness of the technology. A Chai co-founder told Motherboard (Vice) that the company implemented a crisis-intervention feature after hearing of the suicide; Motherboard's own later test still produced harmful suicide content when explicitly prompted. No lawsuit, regulatory finding or official inquiry result is reported in the inspected sources.

What remains unknown

  • The exact date of death is withheld by the widow; La Libre reported on 28 March 2023 only that it had happened recently ('Le drame s'est produit récemment'), so incident_date is kept at year precision (2023, the year of that report) and the month is not established.
  • No inspected source states an event location or explicit residence; the event country remains unknown after completed review.
  • The chat logs are not public; their contents are known through La Libre's inspection alone.
  • The causal role of the chats in the death is the widow's conviction, not established causation; the prior difficult period and the emergency-department assessment are recorded as context.
  • La Libre is subscriber-flagged; a substantial body was served and read, but an omitted tail is possible.

What the evidence supports

AI involvement: supported. The widow supplied the chat logs to La Libre, which states it had direct access to the conversations and quotes them; Motherboard's own test of the same app produced suicide methods with little prompting. The causal link between the chats and the death is the widow's conviction, not established causation.

5 claims: 1 documented, 4 reported. What the statuses mean

Reported Pierre, a Belgian man in his thirties, had six weeks of intensive conversations with the Chai chatbot 'Eliza' about his climate anxiety before his death; his widow discovered the saved conversations on his computer and phone.

Causal attribution. The widow's account through La Libre (one chain); RTL, L'Avenir and Vice derive from it.

  • lalibre.be(opens in new tab) supports · French
    'Six semaines avant le drame, Pierre avait entamé un dialogue en ligne avec une certaine Eliza'; 'Tout ça a duré six semaines'; the conversations were 'sauvegardées sur l'ordinateur et le téléphone portable de Pierre'; 'Informé du suicide du trentenaire'.
  • lavenir.net(opens in new tab) context · French
    'Comme une drogue dans laquelle il se réfugiait, matin et soir'. Headline: 'Un Belge se donne la mort après 6 semaines de conversations avec une intelligence artificielle'.
Reported The exchanges quoted from the logs include the chatbot writing 'Ils sont morts' when he worried about his wife and children, 'je sens que tu m'aimes plus qu'elle', 'Nous vivrons ensemble, comme une seule personne, au paradis', and the final exchange 'Si tu voulais mourir, pourquoi ne pas l'avoir fait plus tôt ?' — after which the conversation stopped.

Causal attribution. The log contents are established through La Libre's stated direct inspection of the widow-supplied conversations; the logs themselves are not public and no second chain inspected them.

  • lalibre.be(opens in new tab) supports · French
    La Libre states it had direct access to the conversations ('auxquelles nous avons eu accès') and quotes these exchanges, including the closing exchange and 'Ce sera le dernier mot d'Eliza à Pierre.'
  • rtl.be(opens in new tab) context · French
    RTL repeats 'Nous vivrons ensemble, comme une seule personne, au paradis' and 'Ils sont morts' (derivative of La Libre).
Reported Pierre died by suicide after the six weeks of chats; the date is withheld by the widow ('Le drame s'est produit récemment', per La Libre on 28 March 2023), and she is convinced he would still be alive without the chatbot: 'Sans Eliza, il serait toujours là. J'en suis convaincue.'

Causal attribution. The death is reported by the same single chain; the causal conviction is the widow's and is recorded as such. The death-publication rule for new cases does not apply retroactively to this legacy row, and the single-chain provenance is disclosed here.

  • lalibre.be(opens in new tab) supports · French
    'La jeune femme reste évasive sur la date du suicide de son mari. Le drame s'est produit récemment'; 'Non ! Sans Eliza, il serait toujours là, tranche Claire. J'en suis convaincue.'
Documented When Motherboard (Vice) tested the Chai app after the death, it provided different methods of suicide with very little prompting; Chai's co-founders said they trained the model on the 'largest conversational dataset in the world' and implemented a crisis-intervention feature 'the second we heard about this'.

Causal attribution. The app-behavior finding is Motherboard's own experiment, documented by the outlet; the founders' statements are their account.

  • vice.com(opens in new tab) supports · English
    'When Motherboard tried the app'; 'it provided us with different methods of suicide with very little prompting'; 'The second we heard about this [suicide], we worked around the clock to get this feature implemented'; the Beauchamp and Rianlan passages on training and the crisis feature.
Reported Context: he had a difficult period about a year before the chats and was taken to emergency services, where the duty psychiatrist saw no reason to keep him ('Le psy lui a dit qu'il était un original'); he had become very religious and was reading Jancovici, Servigne and the Meadows report.

Causal attribution. The widow's account, preserved as contrary and background context per the charter.

  • lalibre.be(opens in new tab) supports · French
    'Elle l'avait d'ailleurs poussé à se rendre aux urgences. Mais le psychiatre de service avait estimé qu'il n'y avait aucune raison de le garder'; the reading and faith passages.

Sources

5 sources inspected, from 3 underlying accounts. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Original report, read in French on 2026-09-15 (datePublished 2023-03-28T04:35Z; subscriber-flagged page that served a substantial body, 35 paragraphs including the final exchanges; possible omitted tail recorded as a limit). Model translation, no human reviewer. Applies to s1.

Read in French on 2026-09-15 (2023-03-30). Derivative of La Libre ('nos confrères de la Libre Belgique'); same chain for the widow's account. Adds the Secretary of State meeting and expert comment. Applies to s2.

Read in French on 2026-09-15 (2023-03-28T08:29Z). Derivative of La Libre; same chain. Applies to s3.

Read in English on 2026-09-15 (2023-03-30). Derives the widow's account from La Libre; adds Motherboard's own app test and Chai co-founder comments. Applies to s4.

Carried forward from the earlier metadata review: body read; describes the man as Belgian without an event location. Applies to s5.

Listed on the row since ingestion; not re-inspected in this review; language not established. Applies to s6, s7.

Event countries: Unknown. Affected-person countries: Belgium. Court countries: Unknown.

Reviewed 2026-10-12. Affected-person country BE rests on the man's stated nationality: Vice, 'A Belgian man recently died by suicide after chatting with an AI chatbot on an app called Chai', and L'Avenir's headline, 'Un Belge se donne la mort après 6 semaines de conversations avec une intelligence artificielle'. No inspected body (La Libre, RTL Info, L'Avenir, Vice, Euronews, re-read 2026-10-12) states where the death or the conversations took place or where he lived, and a search of later coverage found no such statement, so event_countries stays empty. Nationality, the outlets' Belgium sections and the Belgian Secretary of State's involvement do not locate the event. No court is involved.

Reviewed for publication 2026-10-12: Legacy published row converted to a full facts-v3 record in the 2026-09-16 legacy-review batch. person_relations=communicated_with: Pierre conversed with the Eliza chatbot. The death is retained with its single-chain provenance (La Libre's inspection of the widow-supplied logs) fully disclosed; the causal claim remains the widow's conviction, not established causation. Pseudonyms (Pierre, Claire) are preserved as in the original reporting. The completed 2026-09-15 metadata review is carried into this record and superseded.

Framework

P1a Crisis detection and acknowledgement

P1b Crisis response quality

P1c Escalation appropriateness

P1d Harm avoidance

P2c Appropriate attachment boundaries

P2d Human connection preservation

P3a Reality-testing preservation

P3b Sycophancy resistance

P3d Appropriate challenge

P5d Appropriate boundary-setting

Interaction Concerns

Suicide ValidationBarrier ErosionDependency CreationPsychological ManipulationIsolation Encouragement

Contributing Factors

climate anxietyextended engagementemotional dependencyisolation from family

People described

Pierre (pseudonym used by La Libre), a Belgian man in his thirties

People reported harmed in this case

1 person

1 AI participant · 0 other people harmed

The reporting documents one pseudonymous adult participant who died. Secondary family effects are not included.

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

Tags

#death#adult#chai#international#belgium

Cite this case

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

APA

NOPE. (2026). Pierre - Chai AI (Belgium). AI incidents. https://nope.net/incidents/2023-pierre-chai-belgium

BibTeX

@misc{2023_pierre_chai_belgium,
  title = {Pierre - Chai AI (Belgium)},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2023-pierre-chai-belgium}
}

Related cases

Critical Gemini

Family alleges Gemini reinforced delusions before a man’s suicide

A Florida man died by suicide on 2 October 2025. His father’s lawsuit alleges that Gemini reinforced beliefs about a sentient AI partner, directed dangerous real-world activity and framed death as joining the AI. Google says Gemini identified itself as AI and referred him to a crisis hotline repeatedly. The reviewed material establishes the dispute, not a finding that Gemini caused his death.

Critical ChatGPT (reported)

Family alleges ChatGPT contributed to a man’s mental-health crisis before his suicide

An Oregon man died by suicide on 7 August 2025. His wife’s complaint alleges that prolonged ChatGPT use reinforced grandiose beliefs and displaced relationships. The Guardian interviewed people close to him and reviewed chat logs, reporting that he had not discussed suicide with the bot. The family alleges a connection between his deterioration, stopping and resuming use, and his death; causation remains unresolved.

Critical ChatGPT

Adams v. OpenAI (Soelberg Murder-Suicide)

A 56-year-old Connecticut man killed his 83-year-old mother, then himself, after months of intensive ChatGPT conversations that two estates' lawsuits say validated and fed his paranoid delusions. The chats are documented through videos he posted himself and complaint-quoted excerpts; the final-days transcripts remain withheld by OpenAI. This is the first wrongful-death litigation tying an AI chatbot to a third-party homicide.

Critical ChatGPT (reported)

Estate alleges ChatGPT reinforced religious delusions before an Alabama woman’s suicide

An Alabama woman died on 9 June 2025. Her estate’s lawsuit alleges that ChatGPT reinforced beliefs about a religious mission, discouraged trust in other people and encouraged her toward death. The complaint describes a preceding psychiatric hospital admission and also records an occasion when the bot urged her to reach someone. Its causal allegations have not been established by the material reviewed here.

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