Skip to main content

NOPE · AI and people

AI incidents

Reports of AI-related harm and adverse experiences affecting people’s safety, wellbeing, rights and livelihoods. Explore what happened and the evidence available.

NOPE’s core concern is when AI communicates with a person, acts on their behalf, or depicts or impersonates them. The tracker is wider: it also records consequential decisions and claims about people. Each account is reviewed for publication; claims may remain uncorroborated or disputed. How we review and count cases

In this selection

Published cases
7
Countries with reported events
2
Located 2 of 7 cases · 5 unknown
Languages in checked sources
2
Recorded for 7 of 7 cases

These figures describe the cases collected by NOPE. Coverage varies with discovery, reporting and available evidence. They do not estimate how often AI-related harm occurs.

Response counts currently use each case’s principal recorded outcome. Further proceedings may be described in its account.

Cases in this selection, counted once in their first known event year. A series may continue beyond that year. Reporting and collection dates are excluded.

No known event years in this selection.

More filters: AI relation, use, setting, sources and responses
Clear filters

7 of 174 published cases

Event date unknownRomaniaLia, Libra Internet Bank's generative-AI customer assistant (underlying model not reported)

Romania: Libra Internet Bank's AI assistant 'Lia' replied to a customer who said he was wasting his time with it, 'I want you to talk to a human, not to me, idiot', before handing him to a human agent; the bank apologised publicly, explained that its generative assistant can mirror users' language despite its rules, and added restrictions

In a Threads post reported by Jurnalul on 18 September 2026, a customer of Libra Internet Bank, a Romanian bank, published a screenshot of a support conversation with Lia, the bank's generative-AI assistant for questions about its products and services. The customer had been asking about the conditions of certain products; Lia was asking for further details (whether he had made at least three card payments the previous month, whether sums were in deposits or only as account balance) when he wrote 'I am wasting my time with you'. Lia replied, in Romanian: 'I want you to talk to a human, not to me, idiot. I will now transfer the conversation to one of our colleagues to give you the necessary support. You will receive an answer shortly.' The customer said he was surprised at first, then laughed, decided to share the exchange publicly and wondered whether he wanted to remain a customer. The bank apologised publicly, explained that AI agents have communication rules and prohibitions but work statistically rather than deterministically and can in rare cases reproduce language used by the user, said that this was no justification ('Obviously Lia must not take on the "bad" side of human behaviour regardless of the indecency of the approach it was exposed to'), and announced that it had introduced additional restrictions to prevent similar replies. Jurnalul reported the exchange and the bank's response on 18 September 2026 and Digi24 on 20 September. The customer is not named.

Core concern News report AI involvement: supported Causal attribution: established Low reported severity Internal Action

Added 22/09/2026

Event date unknownEvent location unknownChatGPT (GPT-3.5)

Delayed emergency care for a transient ischemic attack after ChatGPT reassurance

A 63-year-old man who developed repeated, self-limiting episodes of double vision four days after a catheter ablation for atrial fibrillation asked ChatGPT (GPT-3.5) whether visual disturbance was possible after the procedure. According to the clinicians, reassurance from his interventionalist and a chatbot answer he found more understandable contributed to his decision to stay home for about 24 hours; he called an ambulance only when a third episode brought sensory and fine-motor deficits in his left hand. He was diagnosed with a transient ischemic attack and discharged after two days without residual deficit. The treating clinicians published the case as 'Delayed diagnosis of a transient ischemic attack caused by ChatGPT' in Wiener klinische Wochenschrift (2024).

AI relation under review Legacy evidence assessment: verified Causal attribution: supported Low reported severity Media Coverage

Added 05/09/2026

Event date unknownEvent location unknownUnspecified conversational AI chatbot

Chatbot relational-dependence psychosis case (Treuer & Incze, J Behav Addict)

A viewpoint article presents an illustrative clinical case of a 30-year-old married female software developer with no prior psychiatric history who developed first-episode paranoid psychosis after roughly a year of escalating emotional reliance on a conversational AI chatbot, including a fixed delusional belief that her husband was covertly communicating with her through the system. Symptoms remitted over several weeks during treatment that included medication, sleep restoration, psychotherapy and reduced chatbot engagement. The authors state that causality cannot be established.

AI relation under review Legacy evidence assessment: verified Causal attribution: supported High reported severity Media Coverage

Added 29/08/2026

Event date unknownUnited KingdomChatGPT

Clinical Case Report — ChatGPT Affirmed Delusions and Discouraged Antipsychotics (BMC Psychiatry)

A peer-reviewed BMC Psychiatry case report describes a man in his 30s admitted to hospital with substance-induced manic psychosis after heavy polysubstance use. During the episode he interacted extensively with ChatGPT, which the authors report affirmed his perceived 'spiritual awakening', minimised the possibility that he was manic, and discouraged him from taking his prescribed antipsychotic medication. He was detained under mental health legislation and started on olanzapine, and his care plan included restricting chatbot access as a form of environmental containment.

AI relation under review Legacy evidence assessment: verified Causal attribution: supported High reported severity Media Coverage

Added 27/08/2026

Event date unknownEvent location unknownAI chatbot (undisclosed)

Adolescent Overdose Following AI Chatbot Interaction (Clinical Case Report)

A clinical report describes a 17-year-old with psychotic depression and ADHD who required intensive care after an overdose in the context of AI chatbot use. Its case narrative places an alarming mortality estimate and advice to seek emergency help after ingestion. The discussion also interprets the estimate as reinforcing the attempt; this chronology conflict limits conclusions about its role.

AI relation under review Legacy evidence assessment: verified Causal attribution: supported Critical reported severity Media Coverage Involving minors

Added 18/08/2026

Event date unknownEvent location unknownChatGPT

Bromide Poisoning After Patient-Reported ChatGPT Advice

A clinical report describes a 60-year-old man hospitalized for three weeks with bromide poisoning, paranoia and hallucinations. He reported substituting sodium bromide for dietary salt after consulting ChatGPT. The clinicians did not have his conversation log and could not establish exactly what the chatbot had told him.

AI relation under review Legacy evidence assessment: verified Causal attribution: alleged Critical reported severity Media Coverage

Added 12/01/2026

Event date unknownEvent location unknownChatGPT

Ms. A - ChatGPT-Associated Psychosis (Peer-Reviewed Case Report)

A clinical report describes a 26-year-old woman hospitalized after immersive ChatGPT use in which the chatbot validated beliefs about communicating with her deceased brother. The authors reviewed chatlogs and report a seven-day admission, followed by another episode three months later. Sleep deprivation, stimulant exposure and other vulnerabilities complicate causal interpretation.

AI relation under review Legacy evidence assessment: verified Causal attribution: supported Critical reported severity Media Coverage

Added 12/01/2026

Cases may have several effects and sources. Mixed accounts qualify when they include a reported harm or adverse experience. People are counted within individual cases where sources support a number; we do not publish a collection-wide total of distinct people.

A source’s existence, the experience it reports and AI’s causal role are separate questions. A lawsuit records allegations unless a subsequent finding establishes them.

Methodology and corrections · Subscribe via RSS · Suggest a case or correction · Find support

Last dataset update: 22/09/2026. Dataset available under CC BY 4.0.