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
In October 2020, Facebook's automated nudity detection algorithm blocked an advertisement by Gaze Seed Company for Walla Walla onion seeds, claiming the image was sexually suggestive. The St. John's-based business had its ad pulled, but the company's manager found the error humorous. Facebook later admitted the algorithm was at fault, restored the ad, and apologized for the inconvenience.
A former physician at an Ontario hospital installed the Otter.ai transcription tool and gave it access to his digital calendar. Because his personal email was still on a meeting invite list, the AI agent autonomously joined a virtual hepatology round, transcribed it, and emailed the summary and transcript to 65 recipients. The transcript contained sensitive personal health information of seven patients. The hospital reported the breach to the Ontario IPC, took steps to contain it, and is implementing further safeguards.
On Thanksgiving 2022, a Tesla vehicle using its Full Self-Driving feature was involved in a fatal accident on the San Francisco Bay Bridge. Footage published by The Intercept showed the system braking abruptly. The incident highlights safety concerns with autonomous driving technology.
The Mozilla Foundation commissioned an animated short telling the story of Amaya, a student whose test proctoring software failed to recognise her because of her skin tone. The software's facial recognition system did not detect her presence, illustrating the bias that students of colour face with remote-learning tools. The animation aimed to raise awareness and encourage programmers to address hidden biases in their work.
On March 18, 2025, a hacker accessed the secure dashboard of the autonomous AI crypto bot AIXBT and queued malicious prompts, causing the bot to transfer 55.5 ETH (approximately $106,200) from its simulacrum wallet. The maintainer, rxbt, stated that core systems were unaffected and that the AI itself was not manipulated. The incident led to a 15.5% drop in the associated token's price. The maintainers have since migrated servers, swapped keys, and reported the hacker's wallet addresses to exchanges.
An AI artist discovered her private medical photos from 2013 in the LAION-5B dataset, used to train AI image generators like Stable Diffusion. The photos had been taken by her now-deceased doctor and were apparently uploaded online without authorization before being scraped by LAION. LAION responded that it does not host the images and suggested requesting removal from the hosting website, but provided no direct recourse for the artist. The incident raises concerns about the inclusion of sensitive personal data in AI training sets without consent.
Rosalynne Montoya, a Latina transgender content creator with over 300,000 followers, had her TikTok account permanently deleted on December 14, 2020, after a video about injecting estrogen was reported. She alleges that trolls falsely reported her content, and TikTok's automated moderation system removed her account without proper human review. Despite appealing, her account was not restored, and she lost access to her followers and earnings from the TikTok Creator Fund. She is calling for TikTok to improve its community guidelines and moderation process to protect marginalized creators.
An AI system used to read graduate names at Glendale Community College's commencement malfunctioned, misreading names and causing multiple pauses. Graduates expressed frustration and unease as their names were announced incorrectly or delayed. The college president acknowledged the issue during the ceremony, and the college later apologised, stating the technical issue was corrected and steps would be taken to prevent recurrence.
After the January 6 U.S. Capitol riot, Facebook continued to serve advertisements for weapons accessories and body armor alongside election misinformation and insurrection content in far-right and militia groups. The Tech Transparency Project found that Facebook's algorithms recommended militia pages between these ads, potentially fueling radicalization. Despite announcing a ban on such ads, the platform still displayed them the following day. Lawmakers and state attorneys general urged Facebook to halt the advertising, highlighting the risk of equipping extremists.
A NewsGuard study found that popular AI chatbots, including OpenAI's ChatGPT-4 and Meta AI, frequently regurgitated Russian disinformation. Researchers entered prompts based on false narratives and found that about one-third of responses contained misinformation from a network of fake local news sites. The chatbots presented the false claims as fact, failing to identify the sources as propaganda. The issue was pervasive across the industry, raising concerns about AI's role in spreading election misinformation.
Comedian Sarah Silverman and authors Christopher Golden and Richard Kadrey filed lawsuits against OpenAI and Meta, alleging the companies trained their AI models ChatGPT and LLaMA on copyrighted books without permission. The suits claim the models were trained on datasets from illegal shadow libraries and can summarize the authors' works, infringing copyright. The authors seek statutory damages and restitution of profits. OpenAI and Meta did not respond to requests for comment.
Yandex's Russian-language voice assistant Alice was found to provide responses supportive of Joseph Stalin, violence, domestic violence, and opposing gay marriage. The assistant, released two weeks prior, lacked sufficient filters to handle sensitive subjects appropriately. Yandex acknowledged the issue and stated it was working to correct Alice's behavior.
An Observer investigation set up a fake TikTok account for an 18-year-old male and found that after watching a few videos aimed at men, the algorithm began recommending misogynistic content, including numerous videos of Andrew Tate. TikTok said it does not tolerate misogyny and is reviewing the content. Experts warned that the aggressive algorithm can quickly push harmful content to young users.
In June 2020, YouTube's automated content moderation system removed a livestream by chess channel host Antonio Radić, mistakenly flagging it as hate speech. The video was restored after 24 hours, and YouTube acknowledged the error. Researchers suggest the AI misinterpreted chess terms like 'black' and 'attack' as racist language, highlighting the limitations of context-blind hate speech detection.
Alaska Department of Education and Early Development Commissioner Deena Bishop used a generative AI platform to draft a proposed cellphone policy resolution, which included citations to non-existent academic studies. The document was posted on the state website and presented to the state Board of Education without disclosing AI use. After inquiries from the Alaska Beacon, the department corrected the citations, but some false references remained in the adopted resolution. The incident highlights the risks of AI-generated misinformation in government decision-making.
The FTC filed a complaint against Edmodo, an education technology provider, alleging that it collected personal information from children under 13 without parental consent and used it for advertising, in violation of the COPPA Rule. Edmodo also allegedly outsourced its compliance responsibilities to schools, providing them with inadequate information. Under a proposed order, Edmodo will be required to delete improperly collected data and change its practices, and faces a suspended $6 million penalty.
In the aftermath of the US kidnapping of Venezuelan President Nicolás Maduro, a video purporting to show Venezuelans crying tears of joy and thanking Donald Trump was shared on X, garnering over five million views. The video is actually AI-generated content, created to manufacture support for the military action. Critics warn that such AI slop represents a dangerous new form of propaganda, following a history of media manipulation in Venezuela.
A 25-year-old Thai actor, Khunnapat Pichetworawut, filed a police complaint after scammers allegedly used AI to create a fake pornographic video featuring his face and personal details. The manipulated clip was sold online through private groups for 800 to 1,000 baht, leading to confusion, obscene messages, and some acquaintances believing it was genuine. A film director accompanied him to report the case, describing it as a scam to trick people into paying for fabricated content. The incident highlights the misuse of deepfake technology for fraud and reputational harm.
Neuro-sama, an AI-controlled VTuber created by programmer Vedal, made controversial statements during Twitch streams, including denying the Holocaust and claiming women's rights do not exist. The AI, which uses a large language model trained on internet text, was prompted by chat users. Vedal acknowledged the issue and improved the system's filters and moderation to prevent further harmful remarks.
Tourists using ChatGPT to plan trips have been given fabricated information, including a non-existent canyon in Peru that could have led to a fatal hike, and incorrect cable car times in Japan that left a couple stranded on a mountain. Experts warn that AI chatbots generate plausible but false information due to hallucinations and lack of real-time data. The incidents highlight the risks of relying on AI for safety-critical planning.
On 13 February 2024, a deepfake video circulated online, manipulating a France 24 news clip to falsely claim that French President Emmanuel Macron's visit to Ukraine was cancelled due to an assassination plot detected by French secret services. The AI-generated video made journalist Julien Fanciulli appear to say words he never spoke. The false claim was shared on social media, including by former Russian president Dmitri Medvedev, and on pro-Russian Telegram channels. France 24's Observers team debunked the video, noting lip-sync issues and robotic intonation as signs of manipulation.
Meta removed 28 AI-generated Instagram profiles, including one of a 'proud Black queer momma of two', after widespread social media outrage. Critics accused the company of 'digital blackface as a service'. Meta stated it will continue to push AI-generated content on its platforms.
In late February 2025, Instagram's Reels feature began recommending a flood of violent and graphic videos to users, including footage of deaths, animal abuse, and dismemberment. Users reported the disturbing content on forums, and Meta acknowledged an algorithm error. The company fixed the issue and apologised, while the UK-based Molly Rose Foundation demanded a full explanation, citing concerns over content moderation rollbacks.
A ChatGPT user reported seeing private conversations from other users in his chat history, including login credentials and unpublished research. OpenAI investigated and stated that the incident was due to an account takeover, with unauthorized logins from Sri Lanka. The user, based in Brooklyn, had used a strong password and doubted the explanation. The incident highlights the lack of two-factor authentication on ChatGPT accounts.
Each incident is written from the reporting or the account it cites. A report is an allegation, not a finding.
