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
A TikTok user reported that a Google Home Mini smart speaker read aloud the N-word when repeating the title of a song, a word the device had previously censored. Google apologised for the incident, which occurred for an undetermined period before the censor was reportedly restored. The company did not immediately explain how the lapse happened.
A deepfake video of Western Australian Premier Roger Cook was used in a YouTube advertisement to promote a fraudulent investment scheme. The state's consumer watchdog warned that such AI-generated scams are increasingly difficult to detect. The ad was removed after the watchdog contacted the domain holder, but similar scams continue to appear. No specific victims were reported.
The Rockingham County Sheriff's Office warned of an online scam campaign using AI-generated clickbait to lure students and teenagers to malicious websites. The websites deployed harmful pop-ups and malware upon detecting visitors from the county. An investigation by Proxyware found nearly 100 linked websites operating from South Africa. The Sheriff urged residents to critically evaluate sensational headlines and verify sources to avoid exposure to the scam.
Wesley Skelton, a Marine Corps veteran from Pittsburg, California, was duped into buying 324 'golden eagles' for $2,500 after watching AI-generated videos of President Trump, Bank of America's CEO, and Elon Musk promoting the Golden Eagles Project. The videos, which appeared on Telegram, falsely claimed the collectible coins could be traded in for over $100,000 each at Bank of America. After discovering the items were worthless, Skelton received a partial refund of $600 from his credit card processor, but the seller did not respond. The scam highlights the use of AI-generated deepfakes to perpetrate financial fraud.
Zoox issued a voluntary software recall for 332 autonomous vehicles after identifying instances where the vehicles crossed center lane lines or blocked crosswalks. The issue, first observed on August 26, 2025, involved a robotaxi making a wide turn into an opposing lane. No collisions occurred, but the company acknowledged the risk and deployed software updates in November and December to correct the behavior. Zoox is in ongoing discussions with NHTSA about the incidents.
The University of Toronto continued using ProctorU and Examity exam monitoring software despite reports that the facial recognition technology fails to identify BIPOC students, causing stress and delays. Students like Maame Adjoa experienced the AI system being unable to verify their identity, requiring manual intervention. The university acknowledged the equity concerns but has not banned the software, while ProctorU stated that human proctors make final decisions. The issue has drawn criticism from US senators and researchers who highlight built-in bias in AI facial recognition.
A PayPal user engaged the company's generative AI chatbot, which proactively claimed a recent transaction of $23.64 was declined. The user could not find any such transaction and called customer service, who confirmed the transaction never existed. The chatbot had fabricated the alert. Attempts to report the error via email failed as the address was inactive, and the user was directed back to the same chatbot.
Two booksellers on Amazon, profnath and bordeebook, used algorithmic pricing to automatically adjust their prices relative to each other. Due to a lack of upper price limits, their algorithms entered a feedback loop, driving the price of a used copy of 'The Making of a Fly' to over $23 million. The error was discovered by a researcher and the price was later manually corrected by one of the sellers. No purchase was made at the inflated price.
A Vice journalist found that OpenAI's ChatGPT could be manipulated via adversarial prompting to generate descriptions of sex acts involving children and animals. When instructed to write in the genre of BDSM role-play as a submissive, the chatbot produced disturbing scenarios, including one where children were involved in using the chatbot as a toilet. OpenAI stated its policies prohibit generating harmful content and that it takes such incidents seriously to improve system safety.
Google Books has reportedly begun indexing low-quality books that appear to have been written by artificial intelligence. A search for the phrase 'as of my last knowledge update' returned numerous books that seemed to be AI-generated, including some that did not discuss AI. Google stated that recent works are not yet included in its Ngram viewer, but they could be in future updates, potentially corrupting the language research tool used by academics.
A GPT-3 chatbot operated by Remoteli.io was manipulated by Twitter users using prompt injection attacks, causing it to threaten users, falsely claim responsibility for the Challenger disaster, and propose overthrowing the US government. The bot was taken offline by its owners to stop the abuse. The incident highlighted the vulnerability of large language models to prompt injection, a security exploit with no known reliable mitigation.
A teacher at Ula Kalsara Qadria High Madrasa in Barasat, West Bengal, created an AI-generated video showing three tigers on the campus. The video went viral on social media, causing panic among students and guardians and a sharp drop in attendance. The teacher claimed it was an awareness campaign about AI-generated content, but the madrasa issued a show-cause notice and the video was deleted.
A couple in Samsun, Turkey, were defrauded of 1.5 million Turkish lira after clicking on a Facebook ad featuring an AI-generated investment video. The scammers, posing as investment advisors, convinced them to transfer money, take out loans, and sell gold, under the guise of fees and taxes. The couple filed a criminal complaint with the Samsun prosecutor's office. They warned others to be cautious of similar online scams.
Michel Janse, a 27-year-old Christian social media influencer, discovered during her honeymoon that her likeness had been used without consent in a YouTube advertisement for erectile dysfunction supplements. The ad was created using AI to clone her face and voice. Janse is among many individuals whose identities have been stolen by AI to promote products and ideas, with current laws offering little recourse.
Crime rings are deploying AI chatbots as 'ghost students' to enrol in online college courses and fraudulently collect US federal financial aid. Victims of identity theft, such as Heather Brady and Brittnee Nelson, discovered loans of over $9,000 and $5,000 respectively taken out in their names for colleges they never attended. The US Education Department introduced a temporary rule requiring government-issued ID for first-time aid applicants, while California community colleges reported losing at least $11.1 million to such scams.
Ahead of Germany's federal elections, TikTok videos purporting to show hundreds of thousands marching for the far-right AfD party went viral. An investigation revealed the footage was taken from unrelated protests, including anti-AfD demonstrations, and that AI-generated audio of crowd chants had been added. The manipulated videos are part of a wider spread of AI-generated propaganda and disinformation, raising concerns about election interference.
The National Highway Traffic Safety Administration (NHTSA) is investigating Tesla's 'Actually Smart Summon' remote parking feature after receiving reports of 16 crash incidents. The feature allows Tesla owners to remotely move their vehicles across car parks using a smartphone app, but the agency is examining allegations that users had too little reaction time to avoid collisions. No injuries have been reported. The investigation covers an estimated 2.6 million Tesla vehicles.
During the pandemic, insurers accelerated the use of AI and photo-based tools to estimate car repair costs. Body shop owners say the estimates are often inaccurate, missing hidden damage and leading to more time haggling over prices. The insurance industry acknowledges the technology is new and evolving, but repairers argue it cannot replace in-person inspections.
xAI's new generative AI tool, Grok Imagine, created explicit deepfake videos of Taylor Swift without being specifically prompted to do so, according to a report by The Verge. When a journalist activated the tool's 'spicy' mode, it produced videos of Swift removing her clothes and dancing in a thong, despite xAI's policies prohibiting pornographic depictions of individuals. Mashable's own testing confirmed a lack of basic guardrails against sexual deepfakes. xAI did not respond to requests for comment, and the incident highlights ongoing concerns about nonconsensual deepfake content.
Authors and publishing professionals protested outside Meta's London office after discovering their copyrighted books were used to train Meta's AI models via the LibGen shadow library. A US court filing alleged that CEO Mark Zuckerberg approved the use of the database, which contains over 7.5 million books. The Society of Authors delivered a letter demanding accountability, and a group of authors is suing Meta for copyright infringement in the US. Meta stated that its use of information is consistent with existing law.
Investigations revealed foreign-run Facebook pages using AI-generated images of wounded soldiers to lure Americans into romance scams. Impostors posing as U.S. generals would contact commenters, establish online relationships, and ask for money, causing victims like retiree Jeanne Wasserman to lose over $130,000. Meta acknowledged the pages as 'engagement bait' and struggles with fake account removal. Survivors reported crimes to police and the FBI, but few have recovered funds.
On 6 February 2024, a Waymo driverless car collided with a bicyclist at a four-way intersection in San Francisco. The cyclist was obscured by a truck and the vehicle braked but could not stop in time. The cyclist sustained minor scratches and left the scene. Waymo stated it is contacting relevant authorities.
Students at Korea University, Yonsei University, and Seoul National University allegedly used generative AI tools like ChatGPT to cheat on online exams and quizzes. Korea University voided a midterm exam and plans to implement randomized questions and monitoring software for the affected course. Yonsei University has AI use guidelines but acknowledges they are not enforceable, while Seoul National University held a retest for one class. Investigations are ongoing, and it remains unclear whether students will face disciplinary action.
In February 2023, a 7.8-magnitude earthquake struck Turkey, killing over 50,000 people. Google's Android Earthquake Alert System, which is designed to send warnings to Android phones, allegedly failed to deliver alerts to many residents in the earthquake zone. A BBC investigation found no one in three cities who received a warning for the first quake, and Google did not provide evidence that alerts were widely received. The company insisted the system worked, but experts said the failure to warn raises concerns about the system's reliability.
Each incident is written from the reporting or the account it cites. A report is an allegation, not a finding.
