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
961 incidents
Telegram Channels Allowed Users to Make Non-Consensual Deepfake Porn as Paid Service
Seven channels were connected in a Telegram ecosystem centered around letting subscribers, as a paid service, generate non-consensual deepfake nudes using a bot from submitted photos of women, including underage girls and women who they know in real life.
Facebook Political Ad Delivery Algorithms Inferred Users' Political Alignment, Inhibiting Political Campaigns' Reach
Facebook's political ad delivery system reportedly differentiated the price of user reach based on their inferred political alignment, inhibiting political campaigns' ability to reach voters with diverse political views, which allegedly reinforces political polarization and creates informational filter bubbles.
Amazon’s "Time Off Task" System Made False Assumptions about Workers' Time Management
Amazon’s warehouse worker “time off task" (TOT) tracking system was used to discipline and dismiss workers, falsely assuming workers to have wasted time and failing to account for breaks or equipment issues.
TikTok’s “For You” Algorithm Exposed Young Users to Pro-Eating Disorder Content
TikTok’s young users were allegedly exposed to community-guideline-violating pro-eating disorder content on their algorithmically curated “For You” page that serves videos from any user on its platform.
Sound Intelligence's Aggression Detector Misidentified Innocuous Sounds
Sound Intelligence's "aggression detection" algorithm deployed by schools reportedly contained high rates of false positive, misclassifying laughing, coughing, cheering, and loud discussions.
Facebook Allegedly Failed to Police Hate Speech Content That Contributed to Ethnic Violence in Ethiopia
Facebook allegedly did not adequately remove hate speech, some of which was extremely violent and dehumanizing, on its platform including through automated means, contributing to the violence faced by ethnic communities in Ethiopia.
Workday's AI Tools Allegedly Enabled Employers to Discriminate against Applicants of Protected Groups
Workday's algorithmic screening systems were alleged in a lawsuit allowing employers to discriminate against African-Americans, people over 40, and people with disabilities.
Length of Stay False Diagnosis Cut Off Insurer's Payment for Treatment of Elderly Woman
An elderly Wisconsin woman was algorithmically determined to have a rapid recovery, an output which the insurer based on to cut off payment for her treatment despite medical notes showing her still experiencing debilitating pain.
Algorithm to Distribute Social Welfare Reported for Oversimplifying Economic Vulnerability
Takaful cash transfer program's algorithm which ranks families by their economic vulnerability level to determine financial assistance reportedly oversimplified people's economic situation, fueling social tension and perceptions of unfairness.
Buenos Aires Government Reportedly Used Children's Personal Data in Facial Recognition System for Fugitives
Beginning in April 2019, the Buenos Aires city government reportedly used data from Argentina’s CONARC fugitive database, including children’s identities and reference photos, in its live Facial Recognition System for Fugitives (SRFP). Human Rights Watch found at least 166 children had appeared in CONARC between 2017 and 2020 and warned that the system exposed them to privacy violations and elevated risks of false matches.
Facebook's Ad Delivery Reportedly Excluded Audience along Racial and Gender Lines
Facebook's housing and employment ad delivery process allegedly resulted in skews in exposure for some users along demographic lines such as gender and racial identity.
The Christchurch shooter and YouTube’s radicalization trap
A New Zealand government report released following a right-wing terrorist killing 51 worshippers at two New Zealand mosques which indicated that Youtube's recommendation algorithm played an important role in the terrorist's radicalization.
Model 3 Tesla on Autopilot Crashed into a Truck in Florida, Killing Driver
In Florida, a Model 3 Tesla on Autopilot mode crashed into a tractor-trailer truck, killing the 50-year-old driver.
Tesla on Autopilot Crashed into Trailer Truck in Florida, Killing Driver
A Tesla Model 3 driver switched on Autopilot seconds before the crash into the underbelly of a tractor-trailer on a highway in Florida, killing the Tesla driver.
GPT-2 Reportedly Reproduced Personal Data from Its Training Data
OpenAI's GPT-2 reportedly memorized and reproduced portions of its training data, including personal information such as names, email addresses, social media handles, and phone numbers. Researchers raised concerns that large language models could expose private or sensitive information when trained on web-scale datasets containing personal data.