AI Incident Database
Browse AI incidents
Every incident record (metadata only) from the weekly snapshot of 2026-09-07. Filter, then export the selection with its licence attached; each row links to the full record and its reports.
Filter incidents (1 active)
-
TikTok’s Content Moderation Allegedly Failed to Adequately Take down Videos Promoting Eating Disorders AIID ↗
Videos promoting eating disorders evaded TikTok's automated violation detection system without difficulty via common misspellings of search terms, bypassing its ban of violating hashtags such as "proana" and "anorexia".
-
Robot in Chinese Shopping Mall Fell off the Escalator, Knocking down Passengers AIID ↗
A shopping guide robot deployed by the Fuzhou Zhongfang Marlboro Mall was shown on video allegedly walking to the escalator by itself, falling down, and knocking over passengers, which prompted its suspension.
-
FaceApp Predicted Different Genders for Similar User Photos with Slight Variations AIID ↗
FaceApp’s algorithm was reported by a user to have predicted different genders for two mostly identical facial photos with only a slight difference in eyebrow thickness.
-
Korean Chatbot Luda Reportedly Made Offensive Remarks Towards Minority Groups in South Korea AIID ↗
A Korean interactive chatbot was allegedly shown in screenshots to have used derogatory and bigoted language when asked about lesbians, Black people, and people with disabilities.
-
Frontline workers protest at Stanford after hospital distributed vaccine to administrators AIID ↗
In 2020, Stanford Medical Center's distribution algorithm only designated 7 of 5,000 vaccines to Medical Residents, who are frontline workers regularly exposed to COVID-19.
-
AI incident: sexual content (Dec 2020) AIID ↗
An AI Incident Database record involving sexual content, dated 15 December 2020. Risk classification, related incidents and the laws on record for this risk area.
-
Online Trolls Allegedly Abused TikTok’s Automated Content Reporting System to Discriminate against Marginalized Creators AIID ↗
TikTok's automated content reporting system was allegedly abused by online trolls to intentionally misreport content created by users of marginalized groups.
-
Brand Safety Tech Firms Falsely Claimed Use of AI, Blocking Ads Using Simple Keyword Lists AIID ↗
Brand safety tech firms falsely claimed use of AI, blocking ads using simple keyword lists.
-
Proctoring Algorithm in Online California Bar Exam Flagged an Unusually High Number of Alleged Cheaters AIID ↗
The proctoring algorithm used in a California bar exam cited a third of thousands of applicants as cheaters, resulting in allegations where exam takers were instructed to prove otherwise without seeing their incriminating video evidence.
-
Court Rules Deliveroo Used 'Discriminatory' Algorithm AIID ↗
In December 2020, an Italian court ruled that Deliveroo’s employee ‘reliability’ algorithm illegally discriminated against workers with legitimate reasons for cancelling shifts.
-
Poachers Evaded AI Cameras and Killed Four Rhinos AIID ↗
AI cameras installed by Ezemvelo KZN Wildlife failed to detect poachers when four dehorned rhino carcasses were found.
-
Facebook Mistakenly Blocked Small Business Ads AIID ↗
Facebook’s AI mistakenly blocked advertisements by small and struggling businesses, after the company allegedly leaned more on algorithms to monitor ads on the platform with little review from human moderators.
-
YouTube Recommendation Reportedly Pushed Election Fraud Content to Skeptics Disproportionately AIID ↗
YouTube's recommendation algorithm allegedly pushed 2020's US Presidential Election fraud content to users most skeptical of the election's legitimacy disproportionately compared to least skeptical users.
-
Autonomous Roborace Car Drove Directly into a Wall AIID ↗
An autonomous Roborace car drove itself into a wall in round one of the Season Beta 1.1 race.
-
OpenAI’s GPT-3 Reported as Unviable in Medical Tasks by Healthcare Firm AIID ↗
The French digital care company, Nabla, in researching GPT-3’s capabilities for medical documentation, diagnosis support, and treatment recommendation, found its inconsistency and lack of scientific and medical expertise unviable and risky in healthcare applications. This incident has been downgraded to an issue as it does not meet current ingestion criteria.
-
Meta Allegedly Used Books3, a Dataset of 191,000 Pirated Books, to Train LLaMA AI AIID ↗
Meta and Bloomberg allegedly used Books3, a dataset containing 191,000 pirated books, to train their AI models, including LLaMA and BloombergGPT, without author consent. Lawsuits from authors such as Sarah Silverman and Michael Chabon claim this constitutes copyright infringement. Books3 includes works from major publishers like Penguin Random House and HarperCollins. Meta argues its AI outputs are not "substantially similar" to the original books, but legal challenges continue.
-
AI mistakes referee’s bald head for football — hilarity ensued AIID ↗
In a Scottish soccer match the AI-enabled ball-tracking camera used to livestream the game repeatedly tracked an official’s bald head as though it were the soccer ball.
-
Tesla Autopilot Mistakes Red Letters on Flag for Red Traffic Lights AIID ↗
A Tesla Model 3 misidentified flags with "COOP" written vertically on them as traffic lights.
-
AI Spam Filters Allegedly Block Legitimate Emails Based on Biased Keyword Detection AIID ↗
AlgorithmWatch tested spam filtering algorithms across Gmail, Yahoo, Outlook, GMX, and LaPoste. Their findings reportedly showed that Microsoft Outlook’s spam filter flagged emails based on specific keywords that led to racial and content-based biases blocking legitimate communications. Emails mentioning Nigeria or containing certain financial and sexual health terms were found to be disproportionately marked as spam.
-
#LekkiMassacre: Why Facebook labelled content from October 20 incident ‘false’ AIID ↗
Facebook incorrectly labels content relating to an incident between #EndSARS protestors and the Nigerian army as misinformation.
-
Researchers find evidence of racial, gender, and socioeconomic bias in chest X-ray classifiers AIID ↗
A study by the University of Toronto, the Vector Institute, and MIT showed the input databases that trained AI systems used to classify chest X-rays led the systems to show gender, socioeconomic, and racial biases.
-
Tesla Disabled “Rolling Stop” Functionality Associated with the “Aggressive” Driving Mode AIID ↗
The “rolling stop” functionality within the “Aggressive” Full Self Driving (FSD) profile that was released via a Tesla firmware update was recalled and disabled.
-
AI attempts to ease fear of robots, blurts out it can’t ‘avoid destroying humankind’ AIID ↗
On September 8, 2020, the Guardian published an op-ed generated by OpenAI’s GPT-3 text generating AI that included threats to destroy humankind. This incident has been downgraded to an issue as it does not meet current ingestion criteria.
-
Tiny Changes Let False Claims About COVID-19, Voting Evade Facebook Fact Checks AIID ↗
Avaaz, an international advocacy group, released a review of Facebook's misinformation identifying software showing that the labeling process failed to label 42% of false information posts, most surrounding COVID-19 and the 2020 USA Presidential Election.
-
Ireland's 2020 Leaving Certificate Calculated-Grades Model Reportedly Issued Incorrect Scores to Thousands of Students AIID ↗
After Covid-19 disrupted the 2020 Leaving Certificate exams, Ireland's Department of Education and Skills used a calculated-grades model to issue student results on September 7. The Department later confirmed coding errors in how the model handled Junior Cycle data, resulting in incorrect grades. About 6,100 students received upgraded results, and college-offer eligibility had to be reassessed.
Disclaimer: informational only, not legal advice. Verify every claim against the linked official sources and consult a qualified lawyer before acting.
Frequently asked questions
What is recorded in each incident?
What happened, when, who was involved as developer or deployer, who was harmed, the sectors and countries affected, and the domain and subdomain the harm falls under. Incidents are mirrored from the AI Incident Database and each links back to its source record.
Does an incident mean wrongdoing was proven?
No. These are reported incidents, not findings of liability. Entities named in a report are named as reported, allegations are marked as alleged, and nothing here is a legal determination.
How current is the incident data?
It syncs directly from the AI Incident Database several times a day, so a new incident usually appears within hours. Each record carries the date it was last synced, and the snapshot date is shown on the page.
Does this page include the full incident reports?
No. The report texts are outside the licence this data is shared under, so only the metadata and the classifications appear here. Every row links to the original incident record, which carries the reports themselves.