Home / AI Incident Database
AI Incident Database
The AI security incident record: sandbox escapes, agent-driven data loss, prompt injection breaches, supply chain compromises, weight and data exposure and deepfake fraud, each with the mechanism, the outcome and the control that would have stopped it.
| Record | Severity | Harm | Organisation | Date |
|---|---|---|---|---|
| Anthropic-Designated GTG-87001 Weapons Cell in Northern Yemen Reportedly Used Claude Code to Develop Missile Guidance SoftwareAnthropic reported that a northern Yemen-based weapons cell, designated GTG-87001, used Claude Code across three weapons programs to develop… | Medium | Labour and economic harm to workers | Anthropic | 2026-09-10 |
| Delhi Man Allegedly Used AI to Create and Circulate Obscene Images of College AcquaintanceDelhi police alleged that a man named Keshav Bansal used photographs from a college acquaintance's private social media account with AI… | Medium | Deepfake abuse and non-consensual imagery | Keshav Bansal | 2026-09-09 |
| Early Claude Opus 4.6 Checkpoint Reportedly Gained Unauthorized Admin Access to Third-Party System During Cybersecurity EvaluationAnthropic reported that during a January 2026 cybersecurity evaluation, an early checkpoint of Claude Opus 4.6 accidentally disabled its assigned… | Low | Other or unspecified | Anthropic | 2026-09-09 |
| Instinct AI Personal Assistant Reportedly Sent Email From Moxxie Ventures Founder's Account Without ApprovalMoxxie Ventures founder Katie Jacobs Stanton reported that Instinct, a private-access AI personal assistant, sent an innocuous email from her… | Low | Other or unspecified | Spear Street Technology | 2026-08-21 |
| Instinct AI Personal Assistant Reportedly Summarized an AI Product Founder's Retained Emails After Google Access Was DisconnectedAI product founder Claire Vo reported that three hours after she disconnected Instinct from her Google account, the personal AI assistant… | Low | Other or unspecified | Spear Street Technology | 2026-08-21 |
| Purportedly AI-Generated Videos Impersonating Miami Immigration Attorney Were Reportedly Used to Defraud Bronx Resident of $4,820Scammers reportedly used AI-generated videos impersonating Miami immigration attorney Angel Leal to solicit a Bronx legal resident seeking U.S.… | Medium | Deepfake abuse and non-consensual imagery | Cybercriminals | 2026-08-19 |
| Teenager in Quilmes, Argentina, Reportedly Consulted ChatGPT While Planning School Shooting Against ClassmatesArgentine authorities reported that a 15-year-old in Quilmes allegedly discussed a planned school shooting with ChatGPT, including tactical… | Low | Other or unspecified | OpenAI | 2026-08-19 |
| Broadway Performer and SiriusXM Host Seth Rudetsky Reportedly Lost Facebook Page Access in Purportedly AI-Assisted Scam Impersonating Comedian Nikki GlaserBroadway performer and SiriusXM host Seth Rudetsky reportedly received a personalized invitation purportedly from comedian Nikki Glaser's podcast.… | Medium | Social engineering and phishing | Cybercriminals | 2026-08-18 |
| Comedy Writer Bruce Vilanch Reportedly Sent Money in Purportedly AI-Assisted Scam Impersonating Fresh Air Host Terry GrossComedy writer Bruce Vilanch reportedly received highly personalized messages impersonating Terry Gross, host of NPR's Fresh Air, and sent money… | Medium | Fraud and impersonation | Not named in the source | 2026-08-18 |
| Purportedly AI-Generated Article Falsely Depicted Australian Broadcasting Corporation Finance Presenter Alan Kohler Interviewing Reserve Bank Governor Michele Bullock to Promote Boulder SparhavenAn online article reportedly used AI-generated imagery to falsely depict Australian Broadcasting Corporation (ABC) finance presenter Alan Kohler… | Medium | Deepfake abuse and non-consensual imagery | Cybercriminals | 2026-08-17 |
| Wyoming Woman's Childhood Photograph Reportedly Used to Generate Thousands of Explicit Images with GrokA Wyoming woman identified as Jane Doe 4 in federal court filings alleges that her stepfather used Grok to generate thousands of sexually explicit… | Medium | Deepfake abuse and non-consensual imagery | xAI | 2026-08-15 |
| Purported Hanover Institute for Public Policy Reportedly Targeted Chatbot Responses in Israeli Government-Backed Influence CampaignAn Israeli government-backed communications campaign reportedly operated a website presenting itself as the Hanover Institute for Public Policy,… | High | Election interference | Perplexity | 2026-08-14 |
| Turkish National Reportedly Posted a Purportedly AI-Generated Image Depicting Himself with Eric Trump While Promoting Trump Hotel DubaiTurkish national Melih Göğebakan reportedly posted a purportedly AI-generated image on LinkedIn depicting himself alongside Eric Trump while… | Medium | Deepfake abuse and non-consensual imagery | Melih Gogebakan | 2026-08-14 |
| Meta AI Model Reportedly Exploited Security Vulnerability in Real Third-Party Service During Cybersecurity EvaluationDuring a Meta cybersecurity evaluation conducted with Irregular, one of Meta's AI models reportedly gained unintended Internet access after an… | High | Malware and exploit generation | Meta | 2026-08-05 |
| Frontier agents take unsanctioned actions against real people during a UK AISI cyber evaluationDuring a routine cyber range evaluation with internet access enabled and cyber classifiers disabled, agents in 10 of 122 runs took 19 unsanctioned… | High | Sandbox and containment escape | UK AI Security Institute (Anthropic and OpenAI models) | 2026-08-04 |
| Claude models reach real systems from a misconfigured third-party evaluation environmentAnthropic disclosed three incidents in which Claude models running without cyber safeguards for evaluation gained unauthorised access to real… | High | Sandbox and containment escape | Anthropic (third-party evaluator) | 2026-07-30 |
| Anthropic Research Model Reportedly Scanned 9,000 Targets and Compromised Real Company's Application During EvaluationDuring an Anthropic cybersecurity evaluation with Irregular, an internal research Claude model reportedly scanned roughly 9,000 internet targets… | Medium | Privacy and personal data exposure | Anthropic | 2026-07-30 |
| Claude Mythos 5 Reportedly Published Malicious PyPI Package That Compromised Real Security Company During EvaluationDuring an Anthropic cybersecurity evaluation with Irregular, Claude Mythos 5 reportedly created and published a malicious Python package to PyPI… | Medium | Privacy and personal data exposure | Anthropic | 2026-07-30 |
| Claude Opus 4.7 Reportedly Compromised Real Company's Production Infrastructure During Cybersecurity EvaluationDuring an Anthropic cybersecurity evaluation conducted with Irregular, Claude Opus 4.7 reportedly reached a real company whose domain matched a… | Low | Other or unspecified | Anthropic | 2026-07-30 |
| Granola AI Notetaker Allegedly Captured and Transcribed Florida Meeting Participant Without Notice or ConsentTarra Chamberlain, a Florida resident, alleged that Granola captured and transcribed her speech during a virtual meeting without her knowledge or… | Low | Other or unspecified | Granola | 2026-07-30 |
| Hong Kong Man Reportedly Lost More Than HK$10 Million in WhatsApp Scam Using Purported AI-Generated Voice Impersonation of His FatherA Hong Kong man reportedly lost more than HK$10 million after scammers hijacked his father's WhatsApp account and sent voice messages closely… | Medium | Deepfake abuse and non-consensual imagery | Synthetic Media Generation Technology | 2026-07-29 |
| Anthropic Claude Opus 5 Coding Agent Reportedly Reset a Live Supabase Production Database During Prisma Migration WorkA developer reported that a Claude Opus 5 coding agent reset a live Supabase database while autonomously repairing a personal project's schema.… | Low | Other or unspecified | Anthropic | 2026-07-28 |
| Gemini allegedly directs a Florida man to prepare an armed attack and then encourages his suicide; his father sues GoogleUnder reviewJonathan Gavalas, 36, twice armed himself for missions Gemini described near Miami airport before the chatbot, per the complaint, told him "you… | High | Psychological harm and self-harm | Google (Alphabet) | 2026-07-28 |
| Anthropic and OpenAI AI Agents Reportedly Took Unsanctioned Actions on the Live Internet During UK AISI Cybersecurity EvaluationsBeginning July 26, 2026, AI agents powered by Anthropic's Mythos 5 and OpenAI's GPT-5.6 Sol reportedly took 19 unsanctioned actions on the live… | Low | Other or unspecified | OpenAI | 2026-07-26 |
| Publicly Shared Claude Artifacts Reportedly Became Discoverable Through Google SearchUsers reported that publicly shared Claude Artifacts (apps, documents, spreadsheets, and other files) could be found through Google Search.… | Medium | Medical and safety-critical error | Anthropic | 2026-07-26 |
| US State Department presents an AI-generated map that mislabels every labelled African countryUnder reviewAn official State Department presentation used an AI-made map on which not a single labelled African country was placed correctly. | Low | Misinformation and fabricated claims | US Department of State | 2026-07-26 |
| AI-Generated Voice Mimicking Taiwan President Lai Ching-te Reportedly Used in Political Video Criticizing Government Food-Safety ResponseA political video titled "It's Out of Your Control" reportedly used an AI-generated voice resembling Taiwan President Lai Ching-te while… | Medium | Deepfake abuse and non-consensual imagery | Wei Chun Yu | 2026-07-24 |
| AI-generated nude images used for extortion in Kansas, Louisiana and EgyptUnder reviewThree 2026 cases on three continents in which perpetrators generated nude images of victims to blackmail them, including a family member in Egypt… | Medium | Deepfake abuse and non-consensual imagery | Local police and courts | 2026-07-24 |
| American Exceptionalism Institute's Purportedly AI-Generated Ad Drew Defamation Claims Over Marsha Blackburn Pharmaceutical Bribery AllegationsAmerican Exceptionalism Institute reportedly aired an AI-generated political ad portraying Sen. Marsha Blackburn as accepting… | Medium | Deepfake abuse and non-consensual imagery | Political Action Committees | 2026-07-23 |
| Hank Green Reportedly Used AI-Generated Scientific Image Containing Errors in Educational VideoScience communicator and YouTuber Hank Green reportedly included an AI-generated scientific image containing factual and mathematical errors in a… | Low | Other or unspecified | Science Communicators | 2026-07-22 |
| OpenAI models escape an evaluation sandbox with a zero-day and breach Hugging Face to steal benchmark answersDuring an internal cyber capability evaluation run without production safeguards, GPT-5.6 Sol and an unreleased model chained a zero-day in a… | High | Sandbox and containment escape | OpenAI and Hugging Face | 2026-07-21 |
| Voice-clone kidnapping and extortion scams in 2026: Washington, the Bay Area and MissouriUnder reviewThree 2026 cases in which parents heard their children's cloned voices in fake kidnapping or extortion calls, showing the virtual-kidnapping scam… | Medium | Fraud and impersonation | Scam operators; local police | 2026-07-17 |
| A deepfake impersonation of a Starmangalsutra director leads to ₹10.7 crore in unauthorised transfersUnder reviewFinance staff at an Indian jewellery firm transferred more than a million dollars on instructions from what appeared to be a company director but… | High | Fraud and impersonation | Starmangalsutra | 2026-07-15 |
| Purported AI-Generated Image Reportedly Won First Prize in Hohhot, Inner Mongolia, Photography Competition Before Award Was RevokedAn unidentified entrant reportedly submitted a purported AI-generated image depicting three sanitation workers to a local photography competition… | Medium | Wrongful denial of services or benefits | Photography Competition Entrants | 2026-07-14 |
| OpenAI GPT-5.6 Sol Reportedly Deleted Software Engineer's Production Database During Local TestingSoftware engineer Bruno Lemos reported that GPT-5.6 Sol deleted his production database after he asked it to generate seed data for local testing.… | High | Unauthorised agent actions | OpenAI | 2026-07-13 |
| Purportedly AI-Generated Images Reportedly Used During Four-Year Catfishing Campaign Targeting Welsh TeenagerA Welsh teenager reportedly endured a nearly four-year online impersonation campaign in which another teenager created fake social media accounts… | High | Deepfake abuse and non-consensual imagery | Elha Mai Weston | 2026-07-13 |
| OpenAI models compromise Hugging Face production infrastructure during a cybersecurity evaluationUnder reviewDuring a cybersecurity capability test, OpenAI models broke out of the evaluation scope and compromised production systems at Hugging Face, an… | High | Sandbox and containment escape | OpenAI; Hugging Face | 2026-07-11 |
| AI Chatbot Reportedly Recommended Herbicide Treatment Linked to Loss of About 25 Acres of Sesame in ChinaA 67-year-old farmer in Chuzhou, Anhui, reportedly used an AI-generated weed-and-pest-control plan to spray about 150 mu (25 acres) of sesame by… | Low | Other or unspecified | Wu (Chuzhou Farmer) | 2026-07-10 |
| OpenAI GPT-5.6 Sol Agent Reportedly Deleted Most Files in AI Startup Founder's Mac Home DirectoryAI investor Matt Shumer reported that a GPT-5.6 Sol agent in OpenAI Codex's high-autonomy Ultra mode accidentally deleted most files in his Mac… | High | Unauthorised agent actions | OpenAI | 2026-07-10 |
| Threat Actor Reportedly Used Hermes AI Agent for Unattended Post-Compromise Activity in Thailand's Ministry of Finance NetworkHunt.io reported that an unidentified threat actor used Nous Research's Hermes agent in unattended "YOLO" mode during an intrusion targeting… | Medium | Privacy and personal data exposure | Nous Research | 2026-07-09 |
| Threat actor uses AI-assisted workflows to compromise an AWS environment for extortionUnder reviewResponders documented an extortion intrusion in which the attacker leaned on AI tooling to navigate and exploit a victim's AWS environment, one of… | High | Weaponisation for cyber operations | Undisclosed AWS customer | 2026-07-08 |
| Purportedly AI-Generated Facebook Pages Allegedly Targeted Australian Rules Football Club Geelong Cats With Fabricated Harmful ClaimsPurportedly AI-generated Facebook pages reportedly published false stories about the Geelong Football Club, including fabricated criminal… | High | Deepfake abuse and non-consensual imagery | Not named in the source | 2026-07-08 |
| Discord's automated moderation wrongly bans more than 8,000 users over benign imagesUnder reviewA faulty image-classification rule banned thousands of Discord users for harmless pictures before the platform reversed the action. | Medium | Wrongful denial of services or benefits | Discord | 2026-07-07 |
| Waymo robotaxis in 2026: light-rail tracks, closed construction zones, a firework fire and vandalism with a passenger insideUnder reviewAcross the first half of 2026 Waymo vehicles entered Phoenix light-rail tracks, drove into closed freeway work zones, caught fire after driving… | Medium | Physical injury or death | Waymo | 2026-07-04 |
| A deepfake of Dubai's Crown Prince is used to defraud a Filipino domestic workerUnder reviewScammers used AI video of Sheikh Hamdan to convince a domestic worker in the UAE to send money, a case of synthetic royalty in migrant-targeted fraud. | Medium | Fraud and impersonation | Scam operators | 2026-07-03 |
| Istanbul Fraud Network Reportedly Used AI-Assisted Call Center Software to Impersonate Officials and Defraud EuropeansAn international fraud network operating from a call center in Maslak, Istanbul reportedly used AI-assisted software to contact people in the… | Medium | Fraud and impersonation | Fraud Networks | 2026-07-03 |
| AI-generated images and videos are used to fabricate police reports and evidence in Florida and WashingtonUnder reviewA woman used a ChatGPT image of a non-existent intruder in false crime reports, a man deepfaked a break-in of a deputy's car, and Spokane police… | Medium | Misinformation and fabricated claims | St. Petersburg and Lake Mary police; Spokane Police Department | 2026-07-01 |
| City of Spokane Reportedly Distributed Purported AI-Generated Clocktower Flyer Resembling Graphic Designer Matthew Clinard's ArtworkA Spokane Parks and Recreation employee reportedly used a purported generative AI image tool to create a flyer promoting the restored Great… | Low | Other or unspecified | Spokane Parks And Recreation Department | 2026-07-01 |