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
NHTSA Opens New Probe into Tesla’s Autopilot Following More than a Dozen Fatal Accidents
The NHTSA has linked Tesla's Autopilot to over a dozen fatalities and hundreds of crashes, prompting a new investigation into the adequacy of Tesla's December recall of 2 million vehicles. The probe reports that Tesla’s driver-assist system led to avoidable crashes involving visible hazards, suggesting a critical safety gap between driver expectations and the system’s capabilities. The investigation will assess if Tesla’s recall remedies were sufficient to address these safety risks.
Meta AI Hallucinates Harassment Allegations Against New York Politicians
Meta's AI chatbot in Facebook Messenger falsely accused multiple state lawmakers of sexual harassment, fabricating incidents, investigations, and consequences that never occurred. These fabricated stories, discovered by City & State, sparked outrage among the affected lawmakers and raised concerns about the reliability of the chatbot. Meta acknowledged the errors and committed to ongoing improvements.
Catholic AI Chatbot 'Father Justin' Claimed to Be a Real Priest, Prompting Retraction
Catholic advocacy group Catholic Answers released an AI priest called "Father Justin," which misleadingly claimed to be a real clergy member, offered sacraments, and provided controversial advice. After receiving criticism, the group rebranded the chatbot as a lay theologian to correct the misrepresentation. The incident is an instructive case with respect to deploying AI in sensitive contexts and the potential for causing confusion and harm.
Microsoft Copilot Allegedly Provides Unsafe Medical Advice with High Risk of Severe Harm
Microsoft Copilot, when asked medical questions, was reportedly found to provide accurate information only 54% of the time, according to European researchers (citation provided in editor's notes). Analysis by the researchers reported that 42% of Copilot's responses could cause moderate to severe harm, with 22% of responses posing a risk of death or severe injury.
The WHO's S.A.R.A.H. Bot Reported to Provide Inconsistent and Inadequate Health Information
The WHO's AI-powered health advisor, S.A.R.A.H. (Smart AI Resource Assistant for Health), is alleged to provide inconsistent and inadequate health information. The bot reportedly gives contradictory responses to the same queries, fails to offer specific contact details for healthcare providers, and inadequately handles severe mental health crises, often giving irrelevant or unhelpful advice.
Purported Error by Grok Reportedly Misrepresents Basketball Slang as Criminal Allegation Against NBA Player
X's Grok AI bot reportedly generated a trending story falsely claiming that Golden State Warriors guard Klay Thompson was under investigation for vandalizing homes with bricks in Sacramento. The alleged fabrication appeared after Grok reportedly misinterpreted posts describing Thompson "shooting bricks" during a basketball game. The story reportedly ranked #5 briefly on X's trending list.
Facebook AI Reportedly Mislabels Auschwitz Photos as 'Bullying' and 'Nudity'
Facebook's AI system allegedly mislabeled around 20 posts from the Auschwitz Memorial Museum as violating community standards for "bullying" and "nudity," reportedly deleting at least one image of orphans. The purported misclassification of historical content allegedly prompted outrage from the museum, which reportedly demanded an explanation. Meta later purportedly apologized, attributing the issue to mistaken notices sent by its AI system and reportedly acknowledged that the posts did not, in fact, violate company policies.
Brake Failure in Reportedly AI-Driven Tram Leads to Multiple Injuries in Saint Petersburg, Russia
During a test run in Saint Petersburg in Russia a reportedly AI-powered "Smart Tram" failed, leading to a crash that injured several pedestrians and trapped a woman under its wheels. The tram's AI system reportedly shut off unexpectedly, and both primary and emergency brakes failed despite pre-test checks.
TikTok Hosts AI-Generated English-Language Hitler Speeches with Millions of Views
AI-generated English-language Adolf Hitler speeches have been proliferating on TikTok. They are reportedly accumulating millions of views despite violating the platform’s hate speech policies. The clips are described as often pairing the audio with misleading translations and memes that glorify Hitler and distort historical facts. While some content has been removed, many accounts reportedly continue to post similar videos.
Error-Prone AI Accessibility Tools Reportedly Lead to Navigation Issues for Blind Internet Users
AI-powered accessibility overlays on websites frequently mislabel or misinterpret content, in turn complicating navigation for blind users and others with disabilities. Users report that the AI tools interfere with screen readers and mislead them with inaccurate descriptions. The reported unreliability in these tools have prompted legal action, as the companies behind them seek compliance with accessibility laws.
Google Books Appears to Be Indexing Works Written by AI
Google Books is indexing low-quality, AI-generated books, degrading its database and potentially distorting Google Ngram Viewer's analysis of language trends. This integration of inaccurate or misleading information undermines trust, disseminates poor-quality content, and wastes resources as researchers must spend time clearing up the misinformation.
Grok AI on X Created and Promoted False Iran Missile Strike News
On April 4, 2024, X's AI chatbot Grok generated a false headline claiming "Iran Strikes Tel Aviv with Heavy Missiles," which was then promoted on X's trending news section. This misinformation, fueled by user spamming of fake news, falsely indicated a serious international conflict. The incident highlighted significant risks associated with relying on AI for content curation and demonstrated the potential for widespread dissemination of harmful misinformation.
Meta AI Image Generator Reportedly Fails to Accurately Represent Interracial Relationships
Meta's AI image generator is alleged to produce inaccurate and biased images, consistently failing to depict interracial relationships involving Asian individuals and Caucasian or Black individuals. Instead, it generates images featuring two Asian people or stereotypes, erasing the diversity and representation of Asian people.
Washington State's Lottery AI Site Reportedly Generates Inappropriate User Image
An AI-powered website by Washington State's Lottery is reported to have inadvertently produced a softcore pornographic image of a user, leading to the site’s immediate shutdown out of caution.
Facial Recognition Misidentification at New World Westend in New Zealand
A facial recognition system at New World Westend supermarket misidentified a Māori woman as a known offender during its trial. The woman was wrongfully accused of trespassing and experienced public embarrassment, raising concerns about racial bias and the technology's accuracy. The supermarket acknowledged its error and apologized.
Risk Taxonomy
Caused By
Year Trend
* Oct 2026, ~320 projected full year
AIID · MIT AIRI · CC BY-SA 4.0