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
Amazon's Alexa Reportedly Shows Political Preference Error in Trump-Harris Presidential Race Queries
Amazon's Alexa was found to provide politically biased responses when asked about the 2024 presidential candidates. It refused to give reasons to vote for Donald Trump, citing neutrality, while offering detailed endorsements for Kamala Harris. Amazon labeled the discrepancy an "error" and reportedly corrected it.
Megalopolis Trailer Included Fake AI-Generated Quotes Attributed to Film Critics
Lionsgate pulled the trailer for "Megalopolis" after it was discovered to contain fake quotes from well-known film critics, generated by AI. The quotes falsely criticized Francis Ford Coppola's previous films. The incident was attributed to an error in vetting by marketing consultant Eddie Egan. Lionsgate has since parted ways with Egan and apologized to Coppola and the critics affected by the fabricated content.
Chatbot in Workplace Training at Bunbury Prison Reveals Real Names in Sexual Harassment Case
During workplace training at Bunbury Prison in Western Australia, a trainer used Microsoft's Copilot AI chatbot to generate case study scenarios. The chatbot produced a scenario that included the real name of a former employee involved in a sexual harassment case, revealing sensitive information.
Microsoft Copilot Falsely Accuses Journalist Martin Bernklau of Crimes
Microsoft's Copilot is reported to have falsely accused veteran court reporter Martin Bernklau of committing serious crimes, including child abuse and fraud. The tool is described as having generated defamatory content that not only accused Bernklau of multiple crimes he covered as a journalist but also provided his personal contact details. Attempts by Microsoft to remove the false entries were only temporarily successful, as the defamatory information reportedly reappeared.
ChatGPT Allegedly Defamed Norwegian User by Inventing Child Homicide and Imprisonment
In August 2024, ChatGPT is reported to have falsely claimed that Norwegian citizen Arve Hjalmar Holmen had killed his two sons and been sentenced to 21 years in prison. The fabricated response allegedly included specific details about the supposed crime, despite Holmen never being accused or convicted of any offense. The incident prompted a GDPR complaint for defamation and inaccurate personal data.
Grok AI Chatbot Reportedly Spreads Unfounded Rumors About Trump’s Dentures
Grok, X’s AI-powered chatbot, reportedly spread unsubstantiated claims about former President Trump’s alleged dentures during his interview with Elon Musk. The AI-generated summary is alleged to have falsely stated that Trump’s speech issues were due to missing dentures, despite no evidence. The post was quickly removed, but the incident is an example of concerns over Grok's tendency to amplify misinformation.
Reported Glitch in Waymo Self-Driving Cars Purportedly Triggers Regular All-Night Honking in San Francisco
Waymo self-driving cars in San Francisco's South of Market neighborhood reportedly began honking at each other late at night, disturbing residents' sleep. The autonomous vehicles, reportedly using a parking lot for ride pauses, triggered honking due to a purported glitch in their algorithms. Despite residents' complaints, the issue allegedly persisted for weeks until Waymo acknowledged the problem and began working on a fix.
Apple Intelligence Allegedly Flags Scam Messages as High Priority
Apple’s AI notification summarization and prioritization feature, part of the Apple Intelligence update, has reportedly flagged scam and phishing messages as high-priority, and is allegedly misleading users into trusting fraudulent communications. By rephrasing and condensing messages, the system is allegedly amplifying scam risks inadvertently and reportedly increasing the likelihood of financial harm for users.
TikTok Network Amplifies AI-Generated Nazi Propaganda and Hate Speech
A coordinated neo-Nazi network on TikTok used AI-generated media, including Hitler speeches, to spread Nazi propaganda and extremist content, violating TikTok’s hate speech policies. The network evaded platform moderation through coded language, imagery, and music, with some accounts accumulating millions of views. TikTok’s algorithm further amplified the reach of this content, despite community guidelines.
SearchGPT Reportedly Misleads Users with Incorrect Festival Dates in Demo
OpenAI’s prototype AI tool, SearchGPT, provided incorrect dates for An Appalachian Summer Festival in Boone, North Carolina during a demonstration video. The AI listed dates that were incorrect, potentially misleading users planning to attend the event, but also harming the reputation of OpenAI as the incident occurred during a high-profile event.
AI Chatbots Reportedly Inaccurately Conveyed Real-Time Political News
Over a week of back-to-back, significant breaking political news stories, including the Trump rally shooting and Biden’s campaign withdrawal, AI chatbots reportedly failed to provide accurate real-time updates. Most chatbots gave incorrect or outdated information, demonstrating their current limitations in handling fast-paced news. These incidents suggest the continuing need for improved AI capabilities and caution in their deployment for real-time news dissemination.
False Election Data on Kamala Harris Reportedly Circulated via Grok AI Chatbot
After President Joe Biden stepped aside as a presidential candidate on July 21, 2024, the AI chatbot Grok on X reportedly falsely informed users that Vice President Kamala Harris missed the ballot deadline in nine states. This misinformation, which spread widely on social media, prompted secretaries of state from five U.S. states to urge Elon Musk to address the problem.
Fatalities Reportedly Occur Despite VioGén Algorithm's Low or Negligible Risk Scores
The VioGén algorithm was designed to help Spanish police assess and prioritize the risk of repeat domestic violence incidents. However, its low-risk assessment of Lobna Hemid reportedly led to inadequate protection; her husband murdered her. Since 2007, 247 women have been killed after being assessed by VioGén. A review of 98 homicides found that 55 of the slain women were scored as negligible or low risk.
Gemini AI Allegedly Reads Google Drive Files Without Explicit User Consent
Kevin Bankston, a privacy activist, claims that Google's Gemini AI scans private Google Drive PDFs without explicit user consent. Bankston reports that after using Gemini on one document, the AI continues to access similar files automatically. Google disputes these claims, stating that Gemini requires proactive user activation and operates within privacy-preserving settings.
Grok AI Model Reportedly Fails to Produce Reliable News in Wake of Trump Assassination Attempt
xAI's model Grok, intended to automate news delivery on the X platform, is reported to have struggled to provide accurate information during the attempted assassination of former President Donald Trump. Grok apparently issued incorrect headlines, including false reports about Vice President Kamala Harris being shot and misidentifying the alleged shooter. These errors show the pitfalls of relying on AI for real-time news aggregation, as it allegedly amplified unverified claims and failed to recognize sarcasm, undermining its reliability.