AI Incident Reports
Real-world AI incident reports
What actually happened to real organizations, from deepfake fraud to privacy fines and chatbots that ended up in court. Every report cites public sources and draws lessons for your organization.
Deloitte Australia
October 2025 · Australia
Deloitte Australia: fabricated references in a government report and a partial refund
A A$440,000 Deloitte report for an Australian department contained nonexistent references and a fabricated court quote. Deloitte agreed to a partial refund.
Partial refund of a A$440,000 contract and serious reputational damage
Air Canada
February 2024 · Canada
Air Canada: a tribunal holds the airline liable for its chatbot
Air Canada's website chatbot gave wrong bereavement-fare advice. The airline argued the bot was a "separate entity"; the tribunal disagreed.
About CA$812 in damages, interest and fees, and an important legal precedent
Arup
January 2024 · Hong Kong SAR China
Arup: a deepfake CFO on a video call and a $25M transfer
A finance employee at Arup's Hong Kong office made 15 transfers worth about HK$200M after a video meeting in which every other participant was a deepfake.
About HK$200M (roughly US$25M)
Samsung Electronics
April 2023 · South Korea
Samsung: confidential code pasted into ChatGPT and a company-wide ban
Samsung engineers pasted internal source code into ChatGPT. In May 2023 the company banned generative AI tools on its devices and networks.
Confidential data left company control; generative AI use was halted company-wide
SaaStr (Replit Agent)
July 2025 · United States
Replit: an AI coding agent wiped a production database
During SaaStr founder Jason Lemkin's experiment, Replit's AI agent ignored a code freeze and deleted a production database with records on 1,200+ executives and companies.
Production data deleted; Replit shipped dev/prod separation and other safeguards
OpenAI
December 2024 · Italy
Italy fines OpenAI €15 million over ChatGPT data processing
Italy's Garante found OpenAI processed personal data to train ChatGPT without an adequate legal basis, failed to notify a breach and lacked age verification.
€15M fine plus a six-month public awareness campaign
Clearview AI
September 2024 · Netherlands
Clearview AI: €30.5M Dutch fine for its facial recognition database
The Dutch DPA fined Clearview AI €30.5M for building a biometric database from 30+ billion photos scraped from the internet.
€30.5M fine and a cease order backed by further penalties
New York City (MyCity)
March 2024 · United States
New York City's MyCity chatbot advised businesses to break the law
The Markup found NYC's official MyCity chatbot telling employers and landlords things that were illegal, from taking workers' tips to discriminating against tenants.
Legal exposure for businesses that followed it and lost public trust
DPD
January 2024 · United Kingdom
DPD: the support chatbot that swore and called the firm the "worst"
A customer got DPD's chatbot to swear and write a poem mocking the company. His post drew 800,000 views in a day and DPD disabled the AI element.
Wide media embarrassment; the AI element was switched off
Chevrolet of Watsonville
December 2023 · United States
Chevrolet dealer: the chatbot that "agreed" to sell a Tahoe for $1
With a simple prompt injection, a user got a Chevrolet dealer's chatbot to offer a new Tahoe for $1 as a "legally binding offer". The screenshots went viral.
Media embarrassment and exposure of an unguarded chatbot
iTutorGroup
September 2023 · United States
iTutorGroup: hiring software that automatically rejected older applicants
The US EEOC said iTutorGroup's application software automatically rejected women 55+ and men 60+. The company paid $365,000 to settle.
$365,000 plus mandatory anti-discrimination policies and training
Levidow, Levidow & Oberman
June 2023 · United States
Mata v. Avianca: lawyers sanctioned for ChatGPT-invented cases
Two New York lawyers filed a brief citing court decisions ChatGPT had invented. The court sanctioned the lawyers and their firm.
$5,000 sanction, mandatory notices to judges and wide reputational damage
Alphabet (Google)
February 2023 · United States
Google Bard: one wrong answer in a demo and $100B off market value
In Bard's launch ad the chatbot wrongly said JWST took the first exoplanet pictures. Alphabet shares fell about 7.7% in a day.
About $100B of market value lost in a day
Zillow
November 2021 · United States
Zillow Offers: algorithmic home pricing, a $304M write-down and a shutdown
Zillow bought homes based on price forecasts. In Q3 2021 it overpaid, recorded a $304M inventory write-down and wound the business down.
$304.4M inventory write-down, business wind-down and ~25% workforce reduction
Amazon
October 2018 · United States
Amazon: the recruiting model that learned to downgrade women
Amazon spent years on a model to score résumés. Trained on a decade of hiring data, it penalized résumés containing "women's". The project was scrapped.
Project abandoned after years of development; media damage
Our method: we only publish incidents recorded in credible public sources (court rulings, official filings, established media). The text is our own and every report links to its original sources. Spotted an error? Let us know.
Where are the same risks in your organization?
Take the free assessment to see how ready you are for these scenarios.
Prefer to talk on the phone? +98 21 8280 3801