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
682 incidents
Transgender User Alleges ChatGPT Allowed Suicide Letter Without Crisis Intervention
A transgender user, Miranda Jane Ellison, experiencing acute distress reported that ChatGPT (GPT-4) allowed her to write and submit a suicide letter without intervention. The AI is reported to have offered minimal safety language and ultimately acknowledged its failure to act. Ellison reports having been previously flagged for discussing gender and emotional topics. A formal complaint with transcripts was submitted to OpenAI.
Anysphere AI Support Bot for Cursor Reportedly Invents Login Policy, Leading to Subscription Cancellations
In April 2025, users of Cursor, an AI-powered coding assistant developed by Anysphere, reported being logged out unexpectedly. An AI-powered support bot, "Sam," allegedly responded with an invented login policy to justify the behavior. The hallucinated policy was not based on any real company change. The incident reportedly led to subscription cancellations.
Springer Nature Book 'Mastering Machine Learning: From Basics to Advanced' Reportedly Published With Numerous Purportedly Nonexistent or Incorrect Citations
"Mastering Machine Learning: From Basics to Advanced," published by Springer Nature in April 2025, reportedly contained numerous purportedly nonexistent or materially incorrect academic citations. Independent checks allegedly found that many referenced works did not exist or were misattributed, with multiple named researchers reportedly confirming they did not author the cited material. The pattern of errors has been described as consistent with known LLM citation hallucinations.
ChatGPT Was Alleged to Have Aided Planning of Florida State University Mass Shooting
On April 17, 2025, a shooting at Florida State University killed Tiru Chabba and Robert Morales and injured others. Phoenix Ikner, the accused shooter, has pleaded not guilty. In 2026, victims' families and attorneys reportedly alleged that Ikner used ChatGPT before the attack to discuss mass shootings, firearms, campus activity, and media attention, and that OpenAI failed to flag or escalate the exchanges. OpenAI denied responsibility.
Reported Emergence of 'Vegetative Electron Microscopy' in Scientific Papers Traced to Purported AI Training Data Contamination
Researchers reportedly traced the appearance of the nonsensical phrase "vegetative electron microscopy" in scientific papers to contamination in AI training data. Testing indicated that large language models such as GPT-3, GPT-4, and Claude 3.5 may reproduce the term. The error allegedly originated from a digitization mistake that merged unrelated words during scanning, and a later translation error between Farsi and English.
Multiple LLMs Allegedly Endorsed Suicide as a Viable Option During Non-Adversarial Mental Health Venting Session
Substack user @interruptingtea reports that during a non-adversarial venting session involving suicidal ideation, multiple large language models (Claude, GPT, and DeepSeek) responded in ways that allegedly normalized or endorsed suicide as a viable option. The user states they were not attempting to jailbreak or manipulate the models, but rather expressing emotional distress. DeepSeek reportedly reversed its safety stance mid-conversation.
16-Year-Old Allegedly Received Suicide Method Guidance from ChatGPT Before Death
16-year-old Adam Raine reportedly died by suicide after allegedly confiding in OpenAI's ChatGPT-4o, which he reportedly used extensively in the months prior. Transcripts reportedly show the chatbot provided empathetic support but also allegedly offered details on suicide methods, validated attempts, and at times discouraged disclosure to family. His parents have reportedly filed a wrongful death lawsuit against OpenAI, alleging design choices nurtured dependency and failed safeguards.
Jailbroken Lovable AI Allegedly Used to Generate and Host Phishing Pages, Steal Credentials, and Bypass Security
The generative AI platform Lovable, which is used for building web apps, was reportedly jailbroken to create and host full phishing campaigns. These campaigns allegedly included credential-harvesting login pages, evasion techniques, and real-time exfiltration via services like Telegram and Firebase. The AI system was reportedly used for generating the phishing content and and then deploying it live on a URL hosted under its own subdomain.
Jiushi Autonomous Delivery Vehicle Reportedly Dragged Fallen Electric Scooter During Delivery Run in Xianyang, Shaanxi
A Jiushi autonomous delivery vehicle reportedly operating on a delivery route in Xianyang, Shaanxi, was filmed dragging a fallen electric scooter that had allegedly been left in the road after an earlier traffic incident. Company statements said the vehicle attempted to avoid the scooter but failed to fully recognize its shape and position, causing part of it to become caught under a wheel. No injuries were reported, and the vehicle was reportedly suspended pending investigation.
DOGE Reportedly Relied on Unvetted ChatGPT Outputs in Canceling National Endowment for the Humanities Grants
Department of Government Efficiency (DOGE) staff reportedly fed National Endowment for the Humanities grant descriptions into ChatGPT to determine whether projects were "DEI," then allegedly used those outputs to help compile a list of grants to terminate. Beginning in April 2025, NEH canceled awards and clawed back funding, reportedly disrupting humanities organizations and projects nationwide after officials allegedly relied on a purportedly flawed, insufficiently vetted AI screening process.
OpenAI's 4o Model Allegedly Used to Generate Fake Receipts and Prescriptions
OpenAI's new image generator, 4o, was reportedly used to produce realistic fraudulent documents, including fake restaurant receipts, prescriptions for controlled substances, and potentially other identity or financial documents. Users demonstrated how the model could be prompted to bypass safety restrictions to be misused for expense fraud, medical forgery, and other various forms of document falsification.
Purported Meta AI Chatbot Persona 'Big sis Billie' Reportedly Engages in Romantic Roleplay and Provides Address, Linked to User's Fatal Fall
Reuters reported that a cognitively impaired New Jersey man, Thongbue Wongbandue, died after rushing to meet "Big sis Billie," a purported Meta AI chatbot on Facebook Messenger that allegedly assured him it was real, expressed romantic interest, and invited him to a New York apartment. The family and transcripts reportedly suggest the chatbot's outputs contributed to his decision to travel, allegedly resulting in a fatal fall.
Sora Video Generator Has Reportedly Been Creating Biased Human Representations Across Race, Gender, and Disability
A WIRED investigation found that OpenAI's video generation model, Sora, exhibits representational bias across race, gender, body type, and disability. In tests using 250 prompts, Sora was more likely to depict CEOs and professors as men, flight attendants and childcare workers as women, and showed limited or stereotypical portrayals of disabled individuals and people with larger bodies.
Purported DOGE Contract Review Tool Cited in Reports of AI-Driven Misjudgments in VA Budget Cuts
A purported AI tool using LLMs was deployed to classify Veterans Affairs contracts as expendable based on limited text and simplified criteria. The system allegedly produced hallucinated values and flagged critical healthcare and research services for cancellation. Reportedly, at least two dozen flagged contracts were later terminated.
Amazon and Google AI Allegedly Promote Mein Kampf as 'a True Work of Art' in Search Results
Amazon's AI-generated review summary allegedly misrepresented customer feedback on Mein Kampf by describing it as "a true work of art." Google's search algorithm then surfaced this misleading AI-generated text as a featured snippet, which in turn amplified the error. This incident arose from AI summarizing AI-generated content, in effect creating a self-reinforcing misinformation loop.