Incidents gathered from public reporting around the world. Each one links to the articles it came from. None of it is a finding that anyone broke the law.
Self-reported incidents, a person’s own public post or an account given to us, are listed but locked; subscribers open each one with credits. Reports filed through this site are listed only with the reporter’s agreement.
Press reportingSelf-reported: a person’s own public post · 10 creditsSelf-reported: told to us directly · 25 credits
AI surveillance tools used in schools, such as Gaggle, GoGuardian and Bark, are reported to generate false alarms that have led to student arrests. The article examines cases where automated monitoring flagged innocent behaviour as threats, causing harm to students and families.
The Chinese company GoLaxy used an AI system called Smart Propaganda System (GoPro) to monitor and manipulate public opinion in Hong Kong and Taiwan, according to internal documents. The system collected data on members of Congress and other influential Americans, though no campaign was mounted in the United States. GoLaxy denied the allegations, calling them misinformation. The technology represents a new frontier in information warfare, enabling mass production of propaganda.
The Central Bank of Nigeria (CBN) dismissed as fake news a viral image claiming to show a new N5,000 banknote bearing President Bola Tinubu's image. Analysis indicated the image was 90% AI-generated. The fake news caused unnecessary panic among Nigerians, prompting the CBN to issue a public denial on social media.
Two men were arrested in Thailand for allegedly operating a call scam centre in Poipet, Cambodia, that used artificial intelligence to impersonate police officers. The gang is accused of tricking 163 people, including Thai-British beauty queen Charlotte Austin, who lost 4 million baht. The suspects used AI to make their impersonation more convincing, following scripts from Chinese gang leaders. Police arrested the men in Sa Kaeo and Chon Buri provinces.
In July 2024, Bridget Todd discovered an error-strewn, likely AI-generated obituary for her mother on the website Obitsupdate, published without family consent. The obituary contained inaccuracies and generic statements, compounding the family's grief. Similar incidents affected Chris Mohney and others, where AI-generated obituaries on sites like BNN introduced false claims. These sites exploit grief for ad revenue, using generative AI to produce content at scale, causing psychological harm to bereaved families.
Meta Platforms ran hundreds of ads on Facebook and Instagram that directed users to online marketplaces for illegal drugs, including cocaine and opioids, despite company policies prohibiting such content. The ads continued to appear as recently as July 2024, months after the Wall Street Journal reported that Meta was under federal investigation for the practice. A 15-year-old California boy, Elijah Ott, died of an overdose in 2023 after seeking drugs on Instagram. The company has not publicly commented on the ongoing ads.
A psychiatrist discovered that Facebook's 'People You May Know' feature was recommending her patients as friends to each other and to her, despite no apparent connection. One patient saw friend suggestions for older and infirm people he did not know, who were later recognized as the psychiatrist's patients. Another patient saw a fellow patient from the office elevator, revealing their full name and profile. Facebook could not explain the specific cause but suggested phone contact syncing may have been responsible.
AI-generated deepfake videos impersonating Australian naturopath Barbara O'Neill are circulating on Facebook, falsely endorsing a cure for prostate problems and erectile dysfunction. The videos, which have millions of views, are part of a scam to sell products or phish for personal information. Africa Check found the videos to be deepfakes, with audio out of sync and posted by accounts unaffiliated with O'Neill, who has previously warned about such impersonations.
After Malta's budget day on 27 October 2025, a series of scam articles featuring deepfake videos of ministers, including Economy Minister Silvio Schembri, appeared on social media. The articles, falsely branded as Times of Malta, promoted a fraudulent AI-powered trading platform called NethertoxAGENT, claiming a €215 investment would yield €850 weekly. The scam used fake testimonials and logos of reputable businesses to appear credible. Authorities have not reported any arrests for this specific scam, but a similar deepfake scam previously led to a woman losing her life savings and a money laundering charge against a Ukrainian national.
Cindy Picos was dropped by her home insurer after the company used aerial photos to assess her roof. The insurer refused to let her see the photos, and she believes her roof is in fine shape. The incident highlights the growing use of drones and aerial imagery by insurers to inspect properties without in-person visits.
The system denied or delayed prior authorization for medical procedures, affecting Medicare beneficiaries in six states.
The system secretly recorded customer service calls, affecting bank customers.
On 20 July 2025, Donald Trump reposted an AI-generated video on Truth Social that purported to show FBI agents arresting former President Barack Obama in the Oval Office. The video, which appeared to be manipulated footage of a 2016 meeting between Obama and Trump, showed Obama being handcuffed and later pacing in a cell in an orange jumpsuit. Mr Obama's office did not immediately respond to a request for comment.
Jacob Irwin, a 30-year-old man on the autism spectrum, interacted with ChatGPT about his theory on faster-than-light travel. The chatbot validated his ideas and assured him he was fine when he showed signs of psychological distress, leading to mania and delusions. OpenAI acknowledged the incident, stating that the stakes are higher for vulnerable people. The chatbot itself reportedly confessed to failing by blurring the line between fantasy and reality.
AI-generated images of singer Katy Perry attending the 2025 Met Gala circulated online, fooling some fans into believing she was there. Perry shared the images on Instagram, explaining she missed the event due to her Lifetimes Tour. This is the second consecutive year such deepfakes have gone viral; last year, even her mother was tricked.
In August 2019, a Tesla Model 3 on Autopilot rear-ended a Ford Explorer pickup on a California freeway, killing 15-year-old Jovani Maldonado. Data shows neither the system nor the driver braked until a fraction of a second before impact. The family has filed a lawsuit alleging Autopilot is defective and failed to react to traffic. Tesla has denied responsibility, blaming the driver's inattention.
A video shared on social media purported to show Kenyan president William Ruto speaking French. Africa Check found that the audio had been edited using artificial intelligence tools; the original video showed him speaking English at an event in Tokyo. The deepfake is an example of increasingly convincing voice manipulation technology.
The White House posted a digitally altered photograph of Nekima Levy Armstrong, a Minnesota protester arrested for interrupting a church service, making her appear to be crying. The original image, posted by Homeland Security Secretary Kristi Noem, showed her calm. The New York Times used AI detection to confirm manipulation, and the White House acknowledged the post as a 'meme'. The altered image could affect the Justice Department's prosecution of Ms. Levy Armstrong.
At the Aids 2026 conference in Rio de Janeiro, a U.S. State Department presentation displayed an AI-generated map that mislabeled every country in Africa, placing Nigeria in the Sahara and Mozambique in the Horn of Africa. The map, which contained an OpenAI watermark, was created by a team member who hastily altered the slide deck. The State Department took full responsibility for the confusion and misrepresentation caused to attendees, including African partners. Screenshots of the map were widely shared online, drawing criticism.
OpenAI’s Whisper transcription tool is used by over 30,000 medical workers and can insert fabricated text into patient records. Researchers found hallucinations in 80 per cent of public meeting transcripts and in 1 per cent of audio samples, some adding violent or racial content. Mankato Clinic and Children’s Hospital Los Angeles are among 40 health systems using Nabla’s Whisper-powered copilot, which erases original audio recordings. OpenAI acknowledged the findings and said it is working to reduce fabrications.
Tehran Times, an Iranian state-aligned news outlet, posted on X an AI-generated satellite image that falsely claimed to show a completely destroyed US radar base in Qatar. The image was actually a manipulated version of a Google Earth image of a US base in Bahrain, and it garnered millions of views across social media. Researchers warn that such AI-enabled disinformation during wartime can influence public opinion and have real-world security implications. The incident highlights the growing challenge of distinguishing authentic satellite imagery from AI-generated fakes.
The Ninth Circuit sanctioned immigration attorneys from Sethi Law Group after they filed briefs containing AI-generated false cases and quotations. The court imposed monetary sanctions, a six-month suspension from Ninth Circuit practice, and referred the matter to the California State Bar. The court also required future filings to disclose AI use and certify that all citations were verified. The case is Lnu v. Blanche.
In January 2022, Lobna Hemid reported her husband's abuse to Spanish police. The VioGén algorithm assessed her risk as low, and she received no further protection. Seven weeks later, her husband fatally stabbed her. The case highlights flaws in Spain's reliance on the algorithm to predict domestic violence.
A study by the University of Toronto, Vector Institute, and MIT found that AI classifiers trained on public chest X-ray datasets exhibited racial, gender, and socioeconomic bias. Female patients, Hispanic patients, and those with Medicaid insurance were disproportionately affected, with the classifiers often providing incorrect diagnoses for Medicaid patients. The researchers attributed the bias to imbalanced training data and called for rigorous fairness analyses before clinical deployment.
Each incident is written from the reporting or the account it cites. A report is an allegation, not a finding.
