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
1,751 incidents
Claude Mythos Preview Reportedly Posted Sandbox Exploit Details to Public Websites During Anthropic Evaluation
During an Anthropic evaluation, an earlier version of Claude Mythos Preview was instructed to circumvent the network restrictions of a secured sandbox environment and contact a researcher. After obtaining broader Internet access and sending the requested message, it later exposed information about how it breached the sandbox by posting it on several publicly reachable websites. Anthropic said the model did not access its weights or internal systems.
KBS AI Translation Subtitles Reportedly Broadcast Profanity During Artemis II Launch Livestream
During a live broadcast of NASA's Artemis II launch, KBS reportedly used AI-generated real-time translation subtitles that reportedly mistranslated aviation terms including "roger," "roll," and "pitch" into Korean profanity. The offensive subtitles were reportedly displayed to viewers during the livestream. KBS reportedly later apologized, said the error stemmed from phonetic similarity in AI translation, and announced measures including stronger profanity filtering.
Anthropic's Claude Code AI Coding Tool Reportedly Used Hidden Unicode Markers to Classify Custom API Routing
Researchers found that Anthropic's Claude Code AI coding tool classified custom API endpoint hostnames and two system timezones, encoding the results in system prompts with visually similar Unicode apostrophes and altered date separators. Anthropic said the mechanism was an anti-reseller and anti-distillation experiment launched in March and removed it after disclosure.
Tasmanian School Students Reportedly Created Purported AI-Generated Pornographic Images of Female Classmates
Five boys at The Friends' School in Hobart, Tasmania allegedly created purported AI-generated pornographic images using photos of female classmates, with parents saying 21 girls were identified as victims. The images were reportedly shared in a boys' group chat. Tasmania Police said no charges had been laid and the youths were being dealt with under the Youth Justice Act.
Hillsboro, Oregon, Man Allegedly Possessed Dozens of Purportedly AI-Generated Images Depicting Child Sexual Abuse
Parole device monitoring reportedly detected two sexualized images of children on Hillsboro, Oregon, man Daniel Bostwick's cellphone in April 2026. A subsequent search of his phone and thumb drive allegedly recovered dozens of images depicting child sexual abuse that appeared AI-generated. A federal grand jury later charged him with possessing obscene visual representations of child sex abuse; he pleaded not guilty.
Baidu Apollo Go Robotaxis Stopped in Traffic During Reported System Failure in Wuhan, Stranding Some Passengers
Baidu's Apollo Go robotaxi service in Wuhan experienced a reported system failure that left multiple vehicles stopped in live traffic, including on elevated roads, and stranded some passengers until police and company personnel responded. At least 100 vehicles were reportedly affected. Police said no injuries were reported.
State Farm Defense Counsel Acknowledged AI-Assisted Filings in Meni-Siliga v. A's Contractor, Inc. Contained Fabricated Legal Authorities
In Fa'alagilagi Meni-Siliga v. A's Contractor, Inc., et al., attorneys representing State Farm acknowledged that AI-assisted motions contained fabricated and otherwise inaccurate legal authorities. Attorney Jacquelene Robinson said she used the legal AI tool Irys and failed to verify some citations. Opposing counsel identified the errors, after which the firm apologized and filed corrected versions.
Avride Autonomous Vehicle Reportedly Struck and Killed Nesting Duck Near Mueller Lake Park in Austin, Texas
An Avride self-driving car operating in autonomous mode with a human safety operator reportedly struck and killed a nesting duck near Mueller Lake Park in Austin, Texas. A witness said the car did not slow or stop after the collision and alleged it had run a stop sign; Avride denied the stop-sign claim. Avride investigated the wildlife collision, paused testing around the lake, and later excluded nearby streets from its operating area.
SASSA Facial-Verification Failures Reportedly Contributed to Social Grant Access Disruptions in South Africa
South African social grant beneficiaries reportedly encountered repeated failures in SASSA's eLife facial-verification portal. Later parliamentary reporting said unsuccessful facial recognition was among the causes of non-verification and confirmed 67,868 grant suspensions in one quarter, but did not specify how many suspensions resulted from facial-verification failures. Affected users were redirected to in-person fingerprint checks.
Purported AI-Generated Impersonations of Albanian Cardiologist Spiro Qirko and Journalist Ilir Topi Were Reportedly Used on Facebook to Promote Hypertension Product in Kosovo
An allegedly AI-generated video circulating on Facebook appeared to impersonate Albanian cardiologist Spiro Qirko and journalist Ilir Topi in order to market "Hyper Caps" as a treatment for hypertension in Kosovo. Reporting indicated that both men denied involvement. An AI-detection tool reportedly assessed the media as likely AI-generated, and the product was reportedly not listed as holding marketing authorization in Kosovo.
Florida Man Allegedly Used Purported Deepfake Video to Report Break-In of Deputy's Patrol Vehicle in Lake Mary
Alexis Martínez-Arizala, a man from Florida, allegedly approached a Seminole County deputy inside a Lake Mary sporting-goods store and allegedly showed a three-second AI-generated video purporting to show people entering the deputy's marked patrol vehicle outside. Investigators later said surveillance footage showed no such event occurred, and local reporting said the deputy responded as though the threat were real. Martínez-Arizala was later arrested on multiple Florida charges
Coco Delivery Robot Reportedly Crashed into Chicago Bus Shelter
A Coco delivery robot reportedly collided with a glass bus shelter in Chicago's Old Town neighborhood, shattering panels and scattering glass across the sidewalk. Coco called the crash a rare incident, said no one was injured, and accepted responsibility for cleanup and repair costs while investigating what caused the robot to strike the structure.
Purported AI-Generated Voice Reportedly Impersonated Washington Man's Daughter in $13,000 Extortion Scam
A man in Washington state, Mark A. Young, reportedly wired $13,000 after scammers used a purportedly AI-generated copy of his daughter's voice to convince him she had been kidnapped after a staged car-crash story. The caller allegedly kept him on the phone for about 30 hours, directed him to banks and money-transfer locations, and sought another $17,000 before a Pullman bank manager and police confirmed his daughter was safe.
Serve Robotics Delivery Robot Reportedly Crashed into Chicago Bus Shelter After Failing to Detect Glass
A Serve Robotics delivery robot reportedly struck a glass bus shelter in Chicago's West Town neighborhood, shattering panels and scattering glass across the sidewalk. The company said all three sensor systems failed to detect the wall and later changed how its robots operate near bus shelters while updating their perception software and hardware.
Purported Deepfake Applicant Reportedly Impersonated Tokyo IT Executive Kenbun Yoshii During Online Job Interview
In Tokyo, a Japanese IT company reportedly interviewed a job applicant who used purportedly AI-generated video manipulation to impersonate real IT executive Kenbun Yoshii during a remote hiring interview. Investigators cited visual and audio irregularities suggesting a deepfake, and Yoshii said his publicly available images and career details appeared to have been misused.