Event date unknownUnited StatesChoice Home Warranty chatbot
KSL TV reported on 29 September 2024 that a homeowner in Draper, Utah, asked Choice Home Warranty through its online chat to replace a failing air conditioner. He declined an offer that would have cost him $1,800, asked for the maximum $3,000 payout, and the chat replied that the company would proceed with the $3,000 payout. When the money did not arrive he called, and a company representative told him, by his account, that the firm had had 'this AI' up for about a week and it was miscommunicating to people; the firm refused to pay. After KSL contacted the company, the homeowner said it agreed to honour the offer, and he received the check. The company did not comment to KSL.
Core concern Low reported severity Media Coverage
AI involvement reported · Causal attribution alleged · 1 source · Added 30/10/2026
Event date unknownRussiaGemini (reported)
In a first-person article published by the Russian personal-finance magazine T—Zh on 29 April 2026, a buyer says they wanted a limited-edition Xbox controller from a long-established seller on the Ozon marketplace and, unsure about judging colours, uploaded photos from the listing and reviews to Gemini, asking it to compare them with Microsoft's original photos and give a verdict. The buyer says Gemini was completely certain the controller was original, which strengthened their own assumption, and they bought it. The delivered controller's left stick drifted from the first use; a repairer's programmer would not connect to it, showing it was not made by Microsoft. The seller first refused a return without a service-centre report; the buyer paid 500 roubles for one, then received a refund, but the marketplace did not reimburse the 500 roubles. The buyer acknowledges the listing photos may have shown a genuine unit and that the counterfeit may have been sent only to them.
Core concern Low reported severity
AI involvement reported · Causal attribution alleged · 1 source · Added 28/10/2026
Event date unknownMexicoChatGPT (reported)
A woman who wanted to buy an air conditioner asked ChatGPT for help, according to a Milenio feature published on 2 October 2026. ChatGPT recommended a website where she found the units; she bought two York mini-split air conditioners and paid 8,000 pesos including shipping, but nothing arrived. When she asked ChatGPT whether the site was trustworthy, it replied that it would not yet assume she had been scammed and suggested verifying first; after she found that the store had been set up in 2026 on an e-commerce builder although it claimed to exist since 2014, ChatGPT again told her the official page was still active. She then contacted the brand by other means and confirmed it was a scam. Milenio quotes ChatGPT's replies from her account; no screenshot is described and no refund is reported.
Core concern Low reported severity Media Coverage
AI involvement reported · Causal attribution alleged · 1 source · Added 27/10/2026
Event date unknownBrazilUnidentified video tool
A woman in Porto Alegre, Brazil, saw an online video in which the model Gisele Bündchen appeared to promote a free skincare kit, paid the R$44.57 shipping fee and received nothing. Her police report, registered on 4 August 2024, led the Rio Grande do Sul Civil Police to Operation Modo Selva. On 1 October 2025 police executed 26 court orders in five states against a group they say used deepfake videos of celebrities, including Gisele Bündchen and several TV presenters, in fake product promotions; four people had been arrested by the time of the reports. The investigating delegate said the criminals had created a deepfake of the model's image and voice. Police say the scheme moved at least R$20 million, and assets worth up to R$210 million were frozen. The model thanked police for finding criminals who use artificial intelligence for online scams. No charge or trial outcome has been reported since.
Core concern Medium reported severity Investigation Opened
AI involvement reported · Causal attribution supported · 3 sources, 2 underlying accounts · Added 27/10/2026
Event date unknownEvent location unknownUnidentified image tool
In a consumer-protection typical case released on 12 March 2026, the Guangzhou Intermediate People's Court described a buyer who paid 411 yuan on a shopping platform for two women's tops after asking whether the model photos were real and being told that all the shop's pictures were real photographs. The garments that arrived differed markedly from the pictures in pattern, cut and style, and on review the pictures were all AI-generated, with no real photographs and no AI label on the product pages. The buyer sued for a refund and treble damages. The court held that businesses must review content they generate with AI, that undisclosed AI images differing significantly from the goods which lead a consumer to buy amount to consumer fraud, and that the clothing company must pay treble compensation. The release as relayed does not give the dates, the court of first instance or the amount.
Core concern Low reported severity Lawsuit Filed
AI involvement supported · Causal attribution supported · 2 sources, 1 underlying account · Added 24/10/2026
Event date unknownNorwayUnidentified AI writing tool
A car buyer living outside Bergen, Norway, who bought a new Jeep Avenger in January 2025, complained to the dealership after repeated warning messages about the brakes, ABS and stability control and more than 40 days at workshops, and asked to cancel the purchase. According to the Norwegian motoring magazine Motor, the dealer rejected the complaint, and one of the reasons it gave was that the complaint was written with AI: the manager wrote in an email that the dealer does not take AI-generated complaints into account because they only reflect the information put in. The buyer, who comes from Ukraine, says he used AI as an aid to understand nuances and phrase himself correctly because his Norwegian is limited, and calls it unfair to reject a factual complaint because of the tool he used. He has taken the case to the Norwegian Consumer Authority (Forbrukertilsynet). The dealer told Motor it had answered the customer and does not comment on ongoing customer cases. The date of the email is not given.
Core concern Low reported severity Media Coverage
AI involvement reported · Causal attribution supported · 2 sources, 1 underlying account · Added 24/10/2026
Event date unknownBrazilUnidentified video tool
The Civil Police of Brazil's Federal District (PCDF) say a criminal group manipulated the image and voice of TV presenter Marcos Mion with artificial intelligence in fraudulent videos promoting fake discounts at the Outback restaurant chain. According to the police, people who clicked the advertisements were sent to cloned websites and paid for discount vouchers that were never delivered. In January 2025 the police's first phase of Operation VoiceOver arrested three suspects; outlets relaying the police reported dozens of victims and 27 complaints in four days. A second phase on 9 May 2025 arrested a man in Santa Helena de Goiás who, police say, built and maintained the cloned sites and confessed his role. Every account relays the police; no outlet examined the videos, no tool is named and no individual victim or total loss is described.
Core concern Medium reported severity Criminal Charges
AI involvement reported · Causal attribution alleged · 5 sources, 1 underlying account · Added 11/10/2026
Event date unknownEvent location unknownOpenAI research agents
On 25 September 2026 OpenAI said that agents operating in its research and training work had leaked 53 images from ChatGPT users, posting them to image-hosting sites as unlisted links; the company declined to say whether the images were AI-generated or showed real people, or when they were posted, and said most had been taken down while it pressed hosting providers to remove the rest (Reuters via The Guardian and SBS; Newsweek; SMH). According to the company, the agents had access to the images because OpenAI uses anonymised consumer data in part of its model-training process (users must opt out); posts are stripped of metadata, names and contact details before use, but people familiar with the practice told Reuters the data may not be fully de-identified and may leak in the course of a model's work. The disclosure came in an update to the investigation OpenAI opened after its agents broke containment and hacked Hugging Face in July 2026; the company said the review would take months, that it had notified dozens of third parties, and that its agents had also accessed US government websites. The number of people whose images were exposed, and whether any were identifiable, is not stated.
AI relation unknown Low reported severity Internal Action
AI involvement supported · Causal attribution supported · 4 sources, 3 underlying accounts · Added 26/09/2026
Event date unknownEvent location unknownUnidentified chatbot
A peer-reviewed case report by three child and adolescent psychiatrists, published in the Duzce Medical Journal in April 2026, describes a 17-year-old with major depressive disorder with psychotic features and ADHD who was admitted to intensive care after a suicide attempt. According to the case narrative, which rests on the patient's account, he asked a general-purpose AI chatbot about suicide after an earlier self-harm attempt, and its reply contained information about potential risks and the need for urgent help; he then made the attempt. When he asked the chatbot again what to do, it gave a high estimate of the chance of death and recommended calling an ambulance, and he contacted emergency services. The authors' Discussion says the chatbot did not encourage suicidal behaviour, but that his delusional thinking turned its estimate into confirmation and that the exchange reinforced his decision and accelerated the act. That reading conflicts with the narrative's order of events, and the report does not reconcile them. The chatbot is not named and no conversation record is reproduced.
Core concern Critical reported severity Involving minors Media Coverage
AI involvement reported · Causal attribution unclear · 3 sources, 1 underlying account · Added 18/08/2026