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
Vanderbilt University's Peabody College sent students a message about the Michigan State University shooting that had been generated by ChatGPT. Students received the message as they grappled with the news of the shooting. The university apologised after the message drew criticism.
A couple riding in a Waymo driverless car in San Francisco felt trapped when a pedestrian attempted to cover the vehicle's sensors. The passengers contacted Waymo support and were advised to remain inside the vehicle. The pedestrian eventually walked away, and the passengers were unharmed. Waymo stated that remaining inside is often the safest course of action during such an event.
The Metropolitan Police's automated facial recognition system had a 98% false positive rate, with only two correct matches, neither of whom were criminals, and no arrests made, according to figures obtained by Big Brother Watch. The system, used at events like Notting Hill Carnival, flagged 35 false positives in one day. Police claimed the figure is misleading due to human checks, but campaigners raised concerns about privacy, racial bias, and lack of oversight.
In April 2018, BuzzFeed and Jordan Peele published a video that used AI to make former US president Barack Obama appear to say statements he never made, including an insult aimed at Donald Trump. The video was a deliberate demonstration intended to warn about the danger of deepfake videos in spreading fake news. The article says the technology is not yet perfect but is expected to become indistinguishable from real footage.
During testing in Jokkmokk, Sweden, a Volvo self-driving XC90 SUV's radar sensors became caked with frozen flakes, rendering the vehicle unable to read the road. The incident highlighted the challenge of snow blindness for autonomous vehicles. Ford engineers later proposed a solution using lidar to create high-definition 3-D maps.
Amateur 'sedition hunters' used the PimEyes facial recognition tool to search for suspected rioters from the U.S. Capitol attack. The tool scans over 900 million online images to find matches. Researchers warn that the tool could be abused for stalking or other invasive purposes. The incident highlights the risks of publicly available facial recognition technology.
Threat actors used Gamma, an AI-powered presentation builder, to host a page that redirected recipients to a fake Microsoft SharePoint login portal. Emails sent from compromised legitimate accounts passed authentication checks, while a Cloudflare Turnstile blocked automated security scanners. An adversary-in-the-middle framework validated credentials in real time and captured session cookies, enabling multi-factor authentication bypass on Microsoft accounts. Abnormal reported the campaign on 15 April 2025.
A vulnerability in Microsoft 365 Copilot allowed users to access files without the access being recorded in audit logs, potentially enabling malicious insiders to exfiltrate data undetected. The flaw, discovered by Pistachio's CTO, was reported to Microsoft in July 2025 and fixed in August, but Microsoft decided not to issue a CVE or notify customers. The vulnerability could be triggered accidentally, meaning many organisations' audit logs may be incomplete. Microsoft classified the issue as 'important' but faced criticism for its lack of transparency.
Getty Images has commenced legal proceedings in the High Court of Justice in London against Stability AI. Getty Images alleges that Stability AI unlawfully copied and processed millions of copyrighted images and associated metadata without a license. Getty Images states it had provided licenses to other technology innovators for AI training, but Stability AI did not seek one. The legal action aims to address the alleged infringement of intellectual property rights.
In late June 2023, dozens of AI-generated books of nonsense appeared on Amazon's Kindle Unlimited young adult romance bestseller list. The books, with titles like 'Apricot bar code architecture', were clearly AI-generated and prompted concerns from authors about lost income. Amazon removed the books from the bestseller list but they remained available for purchase.
The Telegraph found that YouTube was recommending graphic self-harm videos to users as young as 13, prompting criticism from UK health secretary Matt Hancock and the NSPCC. A YouTube spokesperson stated the platform has strict policies against such content and removes flagged violations. The NSPCC called on the government to impose a statutory duty of care on social media platforms to protect children.
Transgender and gender nonconforming travellers in the US allege that Transportation Security Administration full-body scanners, which require officers to select a binary gender, routinely flag their bodies as anomalies, leading to humiliating and invasive pat-downs. Travellers including Olivia, Terra Fox and Allister McGuire describe being pressured to expose their genitals or strip in private rooms after alarms, with one woman saying she was forced to show her penis to female officers who then cleared her. The TSA says it is committed to treating all travellers equally and has introduced new transgender awareness training, but internal complaint data shows a disproportionate number of civil rights complaints relate to screening of transgender people. ProPublica’s review of 174 traveller accounts found that most did not file a formal complaint, often from fear of outing themselves or not knowing how.
The San Francisco SPCA deployed a Knightscope K5 security robot to patrol its campus and surrounding sidewalks, recording 360-degree video to deter vandalism and break-ins. Homeless residents allegedly knocked the robot over and smeared it with BBQ sauce, and public outcry grew that the SPCA was using the robot to discourage homeless people from settling. The SPCA denied targeting homeless people but discontinued the pilot programme with Knightscope last week.
A 57-year-old retired woman in Bengaluru was defrauded of Rs 3.75 crore after scammers used an AI-generated deepfake video of spiritual leader Sadhguru Jaggi Vasudev to promote a fake investment platform called Mirrox. Between February and April 2025, she transferred the money across multiple transactions after being added to a WhatsApp group where fabricated profits were shared to build trust. She realised the fraud when she attempted to withdraw funds and the scammers demanded additional fees, then ceased communication. Police are coordinating with banks to freeze the fraudsters' accounts, but recovery is expected to be challenging.
An independent report found that the Australian government's Vera-2R terrorism risk assessment tool included autism spectrum disorders as a risk factor without empirical evidence. The tool was used to assess terrorist offenders, influencing post-sentence orders including ongoing detention. Despite the critical report being received in May 2020, the federal and NSW governments continued to use the tool. The report concluded the tool was 'extremely poor' at predicting risk.
In June 2021, Google Search displayed Kannada as the answer to the query 'ugliest language in India', offending millions of Kannada speakers. The Karnataka government condemned the result and threatened action, prompting Google to remove the response and apologise. Google explained that the result was generated algorithmically based on web content and did not reflect its opinions.
The Edmonton Police Service used Parabon NanoLabs' Snapshot DNA phenotyping tool to generate a facial composite of a suspect from DNA evidence. The generated image depicted a Black individual, despite the technology's inability to accurately predict age, weight, or skin colour. After releasing the image publicly, the police faced widespread criticism from geneticists and the public, who warned it could lead to racial profiling of the Black community. The police subsequently removed the image and acknowledged they had not adequately considered the risks to marginalized communities.
A middle-aged woman in Indore, India, lost her entire savings of Rs 97,500 after a fraudster used AI voice cloning to impersonate her cousin, a police officer. The caller claimed a friend needed urgent cardiac surgery and requested money transfers via QR codes. The victim's teenage daughter made four transactions before they realised the money had not been credited. Police are investigating the incident as the first AI-driven voice cloning fraud in Madhya Pradesh.
Zillow shut down its Zillow Offers iBuying business after its pricing algorithms, known as Zestimates, consistently overvalued homes. The company purchased thousands of homes at above-market prices and is now selling many at a loss, expecting to lose an estimated $380 million. Zillow attributed the failure to the algorithms' inability to accurately forecast home prices in a volatile market. The company will write off $569 million in home inventory and lay off 25% of its staff.
In April 2024, AI-generated deepfake videos depicting non-existent young female relatives of Marine Le Pen and Marion Maréchal promoting French far-right parties went viral on TikTok, amassing over 2 million views. The videos used juvenile faces of the politicians edited onto stolen influencer content and included racist remarks. The Le Pen family expressed displeasure, and Reconquête! flagged the content to TikTok, which subsequently deleted some accounts. Jordan Bardella called for more AI regulation, while French officials noted that National Rally had opposed a provision banning non-consensual deepfakes.
Hackers tricked Instagram's AI support chatbot into granting them access to other users' accounts by faking their location and requesting email changes. The vulnerability reportedly allowed account takeovers, including a verified account of former US President Barack Obama, though Meta denied world leader accounts were compromised. Meta stated the issue has been resolved and impacted accounts are being secured. The incident highlights risks of AI handling sensitive account recovery without adequate verification.
A Tesla driver in Switzerland recorded his car's Autopilot system misidentifying a horse-drawn carriage as various objects, including a truck spinning into a head-on collision. The incident, which occurred on a highway outside Zurich, was a near miss with no accident. Tesla did not respond to a request for comment. The driver posted the video online, highlighting a potential safety flaw in the system's object detection.
Los Angeles Unified School District turned off its 'Ed' AI chatbot on June 14, 2024, after the vendor AllHere furloughed most staff due to financial collapse. The chatbot, which cost $3 million, was designed to provide students and parents with academic guidance and school information. A former AllHere employee alleged that student data was improperly shared with third parties and processed overseas, raising privacy concerns. The district stated it will ensure privacy protections and plans to eventually restore the chatbot.
Greek Minister of Economy and Finance Kyriakos Pierrakakis and his ministry filed a lawsuit against unknown administrators of a Facebook page. They allege that AI was used to create a deepfake video falsely depicting the minister urging citizens to invest in high-yield programmes. The lawsuit accuses the page of deceptive advertising. The EU AI Act classifies deepfakes as limited risk.
Each incident is written from the reporting or the account it cites. A report is an allegation, not a finding.
