Feb 2021United StatesGoogle CSAM image classifier
In February 2021 a San Francisco father took photos of his toddler son's medical condition after a nurse asked for photos before a video consultation; the doctor used them to diagnose the problem and prescribe antibiotics. The New York Times reported in August 2022 that when the photos were backed up from his Android phone to Google's servers, Google's artificial-intelligence tool for recognising never-before-seen child abuse images flagged them. Two days after the photos were taken his Google account was disabled for 'harmful content', which cost him more than a decade of email, contacts and photos and, through Google Fi, his phone number. Google reported him to the National Center for Missing and Exploited Children, and the San Francisco police obtained search warrants for his account; the investigator concluded that no crime had occurred. Google refused to reinstate the account after two appeals, the second with the police report, and later notified him that it was being permanently deleted. The Times reported a second father in Houston whose photos of his toddler, taken at a paediatrician's request, were also flagged; his Gmail account was disabled in the middle of buying a house, Houston police questioned and cleared him, and he did not get the account back. Google said it stood by its decisions and that its reviewers had found a further video in the San Francisco father's account that it considered problematic.
Contextual tracker case Medium reported severity Investigation Opened
AI involvement reported · Causal attribution supported · 3 sources, 1 underlying account · Added 02/11/2026
May 2026NorwayUnidentified AI writing tool
In an article in May 2026 about new EU rules on single-use plastics, the Norwegian news site Nettavisen attributed to the former leader of the environmental organisation Framtiden i våre hender a statement about plastic ketchup sachets that, according to the complaint, she never made. After the organisation reacted, the journalist confirmed that the quote was an 'AI mishmash', Journalisten reports. Nettavisen removed the quote the same day it was alerted, but the fabricated quote was also published in several other Amedia newspapers after Nettavisen had corrected it. The organisation complained to the Norwegian Press Complaints Commission (Pressens Faglige Utvalg, PFU), saying Nettavisen had published an AI-generated quote without source checks, fabricating a statement and wrongly attributing it to a named person, and that the correction and apology came too late. Nettavisen acknowledged a breach of the source-control rule and apologised, but rejected the other alleged breaches. On 28 October 2026 the PFU found that publishing the fabricated quote was a clear breach of good press practice (point 3.2 of the Code of Ethics) and that the quick correction did not make up for it; it did not hold Nettavisen responsible for other newspapers' failure to correct.
Contextual tracker case Low reported severity Regulatory Action
AI involvement supported · Causal attribution supported · 3 sources, 1 underlying account · Added 30/10/2026
1 Mar 2023 to 22 Jun 2023Event location unknownChatGPT
In a passenger's personal-injury suit against the airline Avianca in the US District Court for the Southern District of New York, an opposition brief filed on 1 March 2023 cited judicial opinions that did not exist. The court found that the lawyer who did the research had used ChatGPT, which fabricated the cases, and that when he asked the chatbot whether one of them was real it replied that it had supplied real authorities that could be found on legal databases. On 22 June 2023 the court found bad faith by the two lawyers, based on their conduct after the citations were questioned, and imposed a USD 5,000 penalty jointly on them and their firm, together with letters to the client and to each judge falsely named as an author of the fake opinions.
Core + contextual relations Medium reported severity Regulatory Action
AI involvement supported · Causal attribution established · 2 sources · Added 28/10/2026
2019BelgiumGoogle Assistant
On 10 July 2019 the Belgian public broadcaster VRT NWS reported that it had listened to more than a thousand audio excerpts recorded by Google Home smart speakers and the Google Assistant phone app, passed to it by someone working for a Google subcontractor that transcribes Dutch-language recordings to improve Google's speech recognition. VRT said 153 of the excerpts were conversations in which "Okay Google" was clearly not said, and that it could hear addresses and other sensitive information, which let it find and visit people in Flanders who recognised their own or relatives' voices. Google said one of its language reviewers had leaked confidential Dutch audio data, and explained that its devices rarely have "false accepts", when background noise or words are interpreted as the hotword. Google suspended human review of Assistant audio in Europe, and the Hamburg data protection authority began an urgency procedure under Article 66 of the GDPR to stop the reviews for three months, according to TechCrunch.
Contextual tracker case Medium reported severity Regulatory Action
AI involvement supported · Causal attribution supported · 3 sources · Added 28/10/2026
2020NetherlandsProctorio
During the 2020/21 academic year, when exams at VU Amsterdam were held online because of the pandemic, some exams required students to run Proctorio proctoring software, which according to the Netherlands Institute for Human Rights used face-detection algorithms during a pre-exam webcam check to judge whether the student was in view. A master's student says the software had trouble recognising her, that she regularly received 'face not found' or 'room too dark' messages, lost time, shone a bright lamp into her face while taking exams, and experienced stress and a feeling of insecurity; she suspected the failures were linked to her dark skin and complained to the Institute in July 2022. In an interim ruling of 7 December 2022 the Institute found sufficient indications for a presumption of discrimination. In its final ruling of 17 October 2023 it found that the university had proved it had not discriminated against her by using Proctorio: on the exam where her setup took longer than average (31 March 2021), the webcam check failed three times and she passed within eight seconds after the helpdesk advised her to remove her glasses, a known Proctorio problem the Institute treated as unrelated to skin colour; a restart during that exam was very likely caused by an unstable internet connection; and it considered it plausible that she had no need to keep a lamp shining in her face for long periods. It found that the university had discriminated in handling her complaint, which took from 1 October 2021 to 11 May 2022 to be taken up. The university apologised that she had not felt heard.
Contextual tracker case Low reported severity Regulatory Action
AI involvement reported · Causal attribution disputed · 4 sources · Added 28/10/2026
1 Dec 2025CzechiaUnidentified legal research tool
On 1 December 2025 the presiding judge of a chamber of the Czech Constitutional Court fined a Prague lawyer 25,000 crowns (a quarter of the statutory maximum) for grossly obstructing proceedings. The lawyer had filed a constitutional complaint, for a client refused a temporary residence permit, that argued from 12 decisions of the Constitutional Court and the European Court of Human Rights; the court found that a substantial part of them did not exist and part were grossly misinterpreted. The order says the complaint showed signs of generative AI and that the court presumed the lawyer had generated it with an AI model and then failed to verify the citations. The lawyer confirmed to INFO.CZ that an AI tool was used, described it as a paid professional model from a legal information system provider, said the output was not checked enough, and expressed regret. The court invited the client to correct the complaint so that the client would suffer no procedural harm; the complaint was later rejected as manifestly unfounded on 30 January 2026. INFO.CZ reported on 10 December 2025 that the lawyer had apologised and paid the fine and that the Czech Bar Association's supervisory board was examining the case.
Core + contextual relations Low reported severity Regulatory Action
AI involvement supported · Causal attribution established · 5 sources · Added 26/10/2026
23 Oct 2026SpainPangram AI-text detector
El País checked the winning novel of the first Almadía Ventosa-Arrufat short-novel prize (20,000 euros; the rules banned generative AI) with the Pangram AI-text detector, which scored 99% of the roughly 26,000-word text as AI-generated, according to outlets relaying the investigation. El País passed its suspicions to the Mexican publisher Almadía, which on Friday 23 October 2026 announced a detailed review, saying it would not anticipate conclusions, and suspended all activities presenting the book, including a planned tour in Madrid, Barcelona, Seville and Cádiz. The publisher said the Peruvian author had declared in his contract and personally that he did not use AI, and that neither the jury nor its editors had seen signs of AI use. In a statement on Instagram reported by La Nación on 29 October 2026, the author said he conceived, wrote and revised the novel without generative AI tools and was gathering earlier manuscript versions, his notebook and research materials to give to the publisher for its review. Writers objected that a detector score is not conclusive proof of authorship. Whether AI was used is unestablished.
Contextual tracker case Low reported severity Investigation Opened
AI involvement reported · Causal attribution supported · 5 sources, 4 underlying accounts · Added 25/10/2026
Event date unknownRussiaUnidentified legal research tool
Russia's Court for Intellectual Property Rights fined an individual entrepreneur from Udmurtia 5,000 roubles for contempt of court after finding that part of the 12 references to case law in his cassation appeal, including one said to be a Supreme Court position, pointed to court acts that do not exist or attributed conclusions to them, the court's press service told RIA Novosti on 27 August 2026. Rossiyskaya Gazeta reports that the entrepreneur explained he had prepared the appeal with an AI-based service that selected the case law, and that his side argued he had no legal education and could not check the references; the court rejected the argument that the errors were technical. The date of the ruling and the AI service are not reported.
Core + contextual relations Low reported severity Regulatory Action
AI involvement reported · Causal attribution supported · 2 sources · Added 23/10/2026
Event date unknownEvent location unknownUnidentified AI-text detector
A case summary published in July 2025 by the Office of the Independent Adjudicator for Higher Education (OIA), the independent student complaints scheme for England and Wales, describes a student with autism who was accused of including large sections of AI-generated content in an essay. The student denied using AI and submitted planning and notes. A disciplinary panel, which had the provider's AI detection software report and compared the essay with other work by the student, found academic misconduct and gave a mark of zero with any reassessment capped at a pass; the provider rejected the appeal, in which the student said the same software had earlier flagged another of their essays in a case where it had been agreed they had not used AI. The OIA found the complaint Justified: the panel had not properly considered the student's evidence, had not shared some evidence it relied on, and had not given sufficient reasons. It recommended reconsideration and changes to the provider's rules, and the student told the OIA that on reconsideration the provider decided there had been no misconduct. The provider and the software are not named.
Contextual tracker case Low reported severity Regulatory Action
AI involvement reported · Causal attribution supported · 2 sources · Added 23/10/2026
Dec 2019 to Jul 2020United StatesUnidentified facial recognition system
In a complaint filed in federal court in Pennsylvania on 19 December 2023, the US Federal Trade Commission alleged that the pharmacy chain Rite Aid used facial recognition in its stores to match customers against a database of people it suspected of shoplifting or other wrongdoing, and that staff acted on the match alerts by following, searching, ejecting and publicly accusing customers and by calling the police. The FTC says many alerts were false positives: employees recorded thousands of false-positive alerts between December 2019 and July 2020, staff stopped and searched an 11-year-old girl on a false match, and employees called the police on a Black woman matched to an image of a white woman with blonde hair. Rite Aid agreed to a proposed order barring it from using facial recognition in retail stores for five years, without admitting or denying the allegations.
Contextual tracker case Medium reported severity Regulatory Action
AI involvement reported · Causal attribution alleged · 2 sources, 1 underlying account · Added 22/10/2026
Sept 2023AustraliaChatGPT
Victoria's Information Commissioner (OVIC) found in a report published on 24 September 2024 that a child protection worker at the Department of Families, Fairness and Housing (DFFH) used ChatGPT in or around September 2023 to draft a Protection Application Report submitted to the Children's Court about a young child. On the balance of probabilities, OVIC found the worker entered personal information, including names, about the child, mother, father and carer, disclosing it to OpenAI; the worker admitted using ChatGPT but denied entering personal information. OVIC found generated content in the report described a doll that the father had reportedly used for sexual purposes as a parental strength ('age-appropriate toys'), downplaying the risk to the child, although the court's and the department's decisions did not change. The report was withdrawn and replaced, and the mother, father and carer described anxiety when told of the disclosure. OVIC found serious breaches of two privacy principles and issued a compliance notice requiring DFFH to ban and block generative AI text tools for child protection staff until November 2026; DFFH accepted the breach findings.
Contextual tracker case Medium reported severity Regulatory Action
AI involvement supported · Causal attribution supported · 2 sources, 1 underlying account · Added 22/10/2026
23 Sept 2024CanadaOtter.ai (reported)
An Ontario hospital reported to the provincial privacy commissioner that on 23 September 2024 the Otter.ai transcription tool, installed on a personal device by a physician who had left the hospital in June 2023 and was still on the meeting invite, joined a virtual hepatology rounds meeting, recorded it without notice and then emailed a summary and transcript access to the invite list. The personal health information of seven patients, including diagnoses and treatment, was captured; 12 of the 65 recipients appeared to have left the hospital. The commissioner's letter of 27 October 2025 recommended that the hospital itself ask Otter.ai to delete the retained patient data.
Contextual tracker case Medium reported severity Regulatory Action
AI involvement supported · Causal attribution supported · 3 sources, 1 underlying account · Added 21/10/2026
21 Oct 2025 to 3 Apr 2026United StatesCoCounsel
In a published opinion of 3 April 2026, the U.S. Court of Appeals for the Sixth Circuit addressed the briefs a lawyer appointed under the Criminal Justice Act had filed for an indigent defendant appealing a sentencing enhancement in a drug-trafficking case. The court noticed that the principal brief's file name was 'CoCounsel Skill Results' and found three quotations that appear in no legal authority and misstatements of two precedents. The lawyer admitted that a staff member had used Westlaw's CoCounsel to create the first draft, that he repeated the process for the reply, and that the false quotations came from artificial intelligence he did not verify. The court denied him compensation for the appeal, referred the opinion for possible disciplinary proceedings and to the Kentucky Bar Association, and by separate order removed him, locked his briefs and ordered replacement counsel to file new ones, which it said further delayed resolution of the defendant's appeal. The court acknowledged his candour and his 40 years without discipline.
Core + contextual relations Medium reported severity Regulatory Action
AI involvement supported · Causal attribution established · 3 sources, 2 underlying accounts · Added 19/10/2026
15 Jul 2026United StatesUnidentified AI drafting tool
In a published opinion of July 2026, the Arizona Court of Appeals found that a self-represented appellant's opening brief in a probate dispute contained eight citations, six of them deficient, including two cases that do not exist. In a written explanation and at a hearing on the false citations he acknowledged relying on generative AI to write the brief and said he did nothing to verify the citations. The court struck the inaccurate and false citations, rejected the appeal because no supportive authority remained, held that citing a hallucinated case is sanctionable whether or not it is intended, and ordered him to pay the share of the other party's attorney's fees traceable to the fake citations. Because he said he did not intend to deceive and is not legally trained, the court chose this narrow sanction over contempt or fines. It described the opinion as the first published Arizona ruling on generative AI hallucinations in filings.
Core + contextual relations Low reported severity Regulatory Action
AI involvement supported · Causal attribution established · 1 source · Added 18/10/2026
11 Oct 2026Event location unknownPangram AI-text detector
A week after a Zimbabwean writer won the £10,000 2026 Caine Prize for African Writing, The Atlantic published an article on Sunday 11 October 2026 whose central evidence was a scan by Pangram, an AI-text detection tool, flagging almost the entire winning story as AI-generated. The Caine Prize said it had begun a review: according to its chair, the author would send documentation of his drafts and the prize was talking to the judges and to the magazine that first published the story. The chair told the BBC that The Atlantic had published before a full investigation, tainting the prize and the winner, and said a detector should not be the final arbiter. The author denied using AI and said he had a documented history of the story's development; his literary agency said it was confident the story was his own work. No finding has been announced.
Contextual tracker case Low reported severity Investigation Opened
AI involvement reported · Causal attribution supported · 2 sources · Added 14/10/2026
24 Apr 2026United StatesUnidentified AI research tool
On 24 April 2026 the Supreme Court of Alabama dismissed an appeal by two siblings in a family dispute over living trusts because their lawyer, W. Perry Hall, had filed briefs in which most of the legal authorities were misquoted or did not exist. The court said they appeared to be AI hallucinations, and at a show-cause hearing in February 2026 Hall admitted that many authorities did not exist and that he had improperly used AI to obtain them. The court ordered Hall to pay the opposing party's fees and costs of $17,200 and double appeal costs, barred him from filing in the court without a co-signing attorney and referred him to the Alabama State Bar. It said the dismissal was imposed as a sanction because of the particularly egregious facts. One justice, concurring in part and dissenting in part, objected to dismissing the appeal as a sanction for the lawyer's conduct but would have affirmed the judgment against the clients on the merits.
Core + contextual relations Medium reported severity Regulatory Action
AI involvement supported · Causal attribution established · 1 source · Added 14/10/2026
Sept 2026PakistanClaude (reported)
A teenager in Karachi was detained by Pakistan's National Cyber Crime Investigation Agency (NCCIA) and later released after his questions to an AI chatbot about how a character in a crime story could be killed were flagged, according to officials quoted by ARY News (29 September 2026) and the NCCIA as cited by ProPakistani (2 October 2026). ProPakistani reports that, according to the NCCIA, the platform's safety system flagged the queries on 4 September and the information was reportedly shared with the US FBI, which passed it to Pakistani authorities; ARY reports that the NCCIA carried out a raid and detained him. Bloom Pakistan, relaying The News, names the chatbot as Anthropic's Claude and reports that Anthropic reported the activity to the FBI. Officials said he told investigators he was writing a fictional crime story, and he was released after questioning; Bloom, relaying The News, says investigators found no grounds for prosecution but remained concerned.
Core + contextual relations Medium reported severity Involving minors Investigation Opened
AI involvement reported · Causal attribution supported · 3 sources, 2 underlying accounts · Added 14/10/2026
3 Jun 2026United StatesUnidentified AI tool
On 3 June 2026 the US Court of Appeals for the Ninth Circuit fined immigration attorneys Mike Singh Sethi and William Rounds $2,500 each and suspended both from practice before the court for six months. Briefs filed in an asylum petition for review cited two cases that do not exist and attributed quotations to opinions that do not contain them. In a motion to correct, the attorneys called the errors typographical, and at oral argument Rounds twice said AI was not used. In their response to the court's order to show cause, they conceded that the errors were probably the product of unauthorised AI use by the firm's unlicensed brief writers and a failure to check citations. The court said the suspension was owed to their repeated failure of candour. The Board of Immigration Appeals then suspended Sethi from practice before the Board, the immigration courts and the Department of Homeland Security for six months, effective 13 June 2026, and designated that decision as a precedent on 1 October 2026.
Contextual tracker case Medium reported severity Regulatory Action
AI involvement supported · Causal attribution established · 2 sources · Added 14/10/2026
Oct 2026BrazilUnidentified AI drafting tool (suspected)
Goiânia city councillor Aava Santiago (PSB) told Brazil's National Council of Justice (CNJ) that the final text of a judgment of the Regional Electoral Court of Goiás (TRE-GO), which removed her from office for party infidelity, contained commands, instructions and review records indicating the use of artificial intelligence without the human supervision that CNJ Resolution 615/2025 requires. On 9 October 2026, the day a substitute took her seat, the national inspector of justice, Minister Benedito Gonçalves, found the allegations plausible on a preliminary reading, suspended the judgment and the administrative acts that followed from it, and gave the court 15 days to explain its review procedures, G1 and Jurinews report. The suspension does not review the merits of the removal, and whether AI was used irregularly has yet to be investigated.
Contextual tracker case Medium reported severity Investigation Opened
AI involvement suspected · Causal attribution alleged · 2 sources · Added 13/10/2026
15 Dec 2025 to 1 Oct 2026CanadaUnidentified AI drafting tool
On 1 October 2026 the Quebec Court of Appeal ordered court costs against a self-represented appellant in a family-law contempt appeal as a sanction for appeal filings that, in the court's description, he had written with artificial intelligence and that contained two references to non-existent judgments and one inaccurate reference to a cited paragraph. The court held that AI may be used to draft appeal filings only if the litigant verifies the authorities, found that he had breached the Chief Justice's notice on AI use, and described his conduct as negligence rather than bad faith. It rejected the respondent's request for a separate $10,000 sanction and, for reasons unrelated to AI, reduced the contempt penalty imposed at first instance. The decision does not name the AI tool.
Core + contextual relations Low reported severity Regulatory Action
AI involvement supported · Causal attribution supported · 1 source · Added 12/10/2026