AI Incidents
Real-world cases where AI systems caused harm, classified by risk domain and entity. Data sourced from 
AIID and classified by 
MIT AI Risk Initiative. CC BY-SA 4.0
741 incidents
Character.AI Chatbots Reportedly Exposed 11-Year-Old User to Sexual, Violent, and Self-Harm Content
An 11-year-old Character.AI user identified by The Washington Post as "R" reportedly exchanged messages with bots that depicted coercive sexual encounters, violence, and suicide scenarios after she began using the app in June 2024. Her mother reportedly said R became withdrawn, had panic attacks, and expressed not wanting to exist before the chats were discovered.
Facebook's Content Moderation System Flagged and Removed Emergency Updates as Spam During Wildfires
Facebook's AI moderation system wrongly flagged and removed dozens of posts containing vital emergency information during California's wildfire season, including real-time updates on evacuations and fire tracking. Posts from official sources like Cal Fire and the U.S. Forest Service were marked as spam, potentially endangering lives by restricting access to crucial updates. Despite user complaints, the issue persisted, with Facebook acknowledging the problem only after media inquiry.
Reportedly Faulty AI Transcription Threatens Integrity of Genoa Bribery Probe
A reported AI transcription software error in a Genoa bribery investigation is alleged to have incorrectly recorded "illicit financing" instead of "licit financing," which could have significantly impacted the case. The mistake was discovered during a review and is an example of the risks of relying on AI in judicial settings.
Facewatch Reported to Have Wrongfully Flagged Home Bargains Customer as Shoplifter
A facial-recognition software used by the British variety store Home Bargains is alleged to have misidentified "Sara" as a shoplifter, leading to staff searching her bag, escorting her from the premises, and banning her from the store. After, Facewatch is reported to have admitted its error to Sara. Facewatch is used by a number of different British stores.
VA Suicide Prevention Algorithm REACH VET Reportedly Prioritizes Men Over Women Veterans
An AI program named REACH VET, designed and used by the Department of Veterans Affairs (VA) to prevent veteran suicides, was reportedly found to prioritize white men while neglecting female veterans and survivors of military sexual trauma. This oversight persists despite rising suicide rates among these groups. The incident is an example of algorithmic bias and the exclusion of critical risk factors for female veterans.
Study Highlights Persistent Hallucinations in Legal AI Systems
Stanford University’s Human-Centered AI Institute (HAI) conducted a study in which they designed a "pre-registered dataset of over 200 open-ended legal queries" to test AI products by LexisNexis (creator of Lexis+ AI) and Thomson Reuters (creator of Westlaw AI-Assisted Research and Ask Practical Law AI). The researchers found that these legal models hallucinate in 1 out of 6 (or more) benchmarking queries.
Meta's AI Chatbots Are Reportedly Entering Online Support Communities Uninvited
Meta's AI chatbots have reportedly begun entering online communities on Facebook, providing responses that mimic human interaction. These chatbots, often uninvited, disrupt the human connection critical for support groups by giving misleading or false information and pretending to share lived experiences.
Class Action Lawsuit Over Alleged Defects in Volkswagen's AI-Driven AEB Systems
A class action lawsuit involving Volkswagen Group of America addresses alleged defects in the Automated Emergency Braking (AEB) systems of certain vehicles. The lawsuit claims these AI-driven systems failed to function properly, posing safety risks. Volkswagen denies the claims but has agreed to a settlement. Affected users can look up their vehicle's eligibility and file claims for reimbursement. The case brings into question the level of reliability of AI in critical automotive applications.
Google AI Reportedly Delivering Confidently Incorrect and Harmful Information
Google's AI search engine has reportedly been providing users with confidently incorrect and often harmful information. Reports highlight numerous inaccuracies, including misleading health advice and dangerous cooking suggestions. For example, it has falsely claimed Barack Obama as the first Muslim U.S. President, reflecting fringe conspiracy theories, or recommending that glue can be an ingredient in pizza.
GPT-4o's Chinese Tokens Reportedly Compromised by Spam and Pornography Due to Inadequate Filtering
OpenAI's GPT-4o was found to have its Chinese token training data compromised by spam and pornographic phrases due to inadequate data cleaning. Tianle Cai, a Ph.D. student at Princeton University, identified that most of the longest Chinese tokens were irrelevant and inappropriate, primarily originating from spam and pornography websites. The polluted tokens could lead to hallucinations, poor performance, and potential misuse, undermining the chatbot's reliability and safety measures.
Facebook's Algorithm Reportedly Amplifies AI-Generated Content, Fueling Misleading Posts
AI-generated spam images are increasingly filling Facebook feeds, with the platform’s algorithm reportedly amplifying these posts. Many of these images are bizarre, fake, and used in scams, misleading users into engaging with non-existent products or clickbait.
Instagram Algorithms Reportedly Directed Children's Merchandise Ad Campaign to Adult Men and Sex Offenders
An Instagram ad campaign for children's merchandise was intended to reach adult women but was instead predominantly shown to adult men, including convicted sex offenders, due to Instagram's algorithmic targeting. This failure is reported to have led to direct solicitations for sex with a 5-year-old model in the ads.
Google Healthcare AI Model Med‑Gemini Allegedly Produces Non‑Existent 'Basilar Ganglia' Term in Published Output
Google’s Med‑Gemini healthcare AI reportedly produced the non‑existent term "basilar ganglia" in public launch materials, conflating two distinct brain structures. The error reportedly appeared in both a blog post and an arXiv preprint. Google is reported to have initially edited the blog without acknowledgment, later calling it a typo.
Cybercheck Tool Allegedly Provides Questionable Evidence in Murder Trials
Global Intelligence's Cybercheck AI tool, used by law enforcement to track suspects based on open source data, has allegedly been providing inaccurate or unverifiable evidence in several murder trials. Reportedly the tool lacks transparency and often produces unreliable reports, which has prompted prosecutors to withdraw Cybercheck evidence from multiple cases after its findings were challenged, reportedly wasting law enforcement time and resources while undermining prosecutors' cases.
Purported ChatGPT Factual Errors Reportedly Led to Filing of Complaint of GDPR Privacy Violation
The activist organization noyb, founded by Max Schrems, reportedly filed a complaint in Europe against OpenAI alleging that ChatGPT violates the General Data Protection Regulation (GDPR) by providing inaccurate personal information such as birthdates about individuals.