AI Incident Database

Source-qualified evaluations, report statistics and safety observations.

Reported incidents by year

Counts describe the retained AIID snapshot. Estimated dates remain included as source assertions; unknown dates remain unknown. No usage denominator or company risk ranking is available.

  • 19831 1
  • 19921 1
  • 19961 1
  • 19981 1
  • 19991 1
  • 20034 4
  • 20061 1
  • 20071 1
  • 20083 3
  • 20092 2
  • 20101 1
  • 20113 3
  • 20128 8
  • 20136 6
  • 201413 13
  • 201524 24
  • 201641 41
  • 201751 51
  • 201846 46
  • 201945 45
  • 202091 91
  • 202180 80
  • 2022109 109
  • 2023179 179
  • 2024313 313
  • 2025466 466
  • 2026309 309

Reported risk domains

  • Malicious Actors & Misuse566 566
  • AI system safety, failures, and limitations313 313
  • Unclassified305 305
  • Discrimination and Toxicity243 243
  • Misinformation196 196
  • Privacy & Security112 112
  • Human-Computer Interaction42 42
  • Socioeconomic & Environmental Harms24 24

Students in Maltese Schools Reportedly Being Targeted by Deepfake Nudes

Beginning sometime in October 2024, students ranging in age from 12 to 20 in schools in Malta were targeted in deepfake nude incidents. The attackers have been superimposing the faces of some of the students over the bodies of naked girls or using "nudify" apps to "undress" the photos, which have reportedly been sent to the victims over WhatsApp in order to humiliate them or in some cases blackmail them.

Malicious Actors & Misuse · Date 2024-10-01 · Estimated Unknown

Reported developer
Deepfake Technology Developers
Reported deployer
Deepfake Creators, Deepfake Creators In Malta
Reported affected
Students In Malta, Students, Minors In Malta, Minors, General Public Of Malta, Educational Communities
Source snapshot ↗

Reported Darknet Launch of Xanthorox AI Introduces Autonomous Cyberattack Platform

Xanthorox AI is a malicious, modular AI system released on darknet forums in early 2025. Designed from scratch for offensive cyber operations, it runs on private infrastructure and includes models for code generation, phishing, malware, social engineering, and real-time voice/image input. Its release represents a deliberate deployment of an autonomous attack platform.

Malicious Actors & Misuse · Date 2025-04-07 · Estimated Unknown

Reported developer
Xanthorox Ai Creators, Unknown Black Hat Ai Developers
Reported deployer
Unknown Malicious Actors, Darknet Forum Users, Cybercriminals
Reported affected
Victims Of Phishing Attacks, Victims Of Malware Attacks, Victims Of Automated Cybercrime, General Public, Enterprise It Systems, Critical Infrastructure Systems
Source snapshot ↗

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.

Malicious Actors & Misuse · Date 2025-04-09 · Estimated Unknown

Reported developer
Loveable Ai
Reported deployer
Unknown Actors, Cybercriminals
Reported affected
General Public, Users Targeted By Phishing Attacks, Impersonated Organizations, Microsoft
Source snapshot ↗

Alleged Deepfake Investment Scam in Spain Defrauds 208 Victims of €19 million ($20.9 million)

Spanish police arrested six individuals allegedly behind a €19M ($20.9M) global investment scam powered by AI. The operation used deepfake ads featuring national celebrities to deceive victims, many of whom were selected through targeting algorithms. Scammers posed as financial advisors and fake officials, cycling through romance baiting, investment fraud, and recovery scams. AI-generated content amplified trust and engagement.

Malicious Actors & Misuse · Date 2025-04-07 · Estimated Unknown

Reported developer
Deepfake Technology Developers, Generative Ai Developers
Reported deployer
Unnamed Criminal Group In Spain, Individuals Arrested In Granada And Alicante
Reported affected
General Public, General Public Of Spain, Global Victims Targeted By Scams, Individuals Who Lost Money Through Scam Investment Platforms
Source snapshot ↗

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.

Malicious Actors & Misuse · Date 2025-03-31 · Estimated Unknown

Reported developer
Openai
Reported deployer
Scammers
Reported affected
Regulatory Agencies, Pharmacists, Law Enforcement, Healthcare Providers, General Public, Expense Management Platforms, Employers
Source snapshot ↗

Alleged AI-Generated Clone of Exante Brokerage Used to Defraud U.S. Investor via JPMorgan Account

Scammers used AI tools to clone the broker Exante and defraud at least one U.S. victim by registering a JPMorgan Chase account and replicating Exante’s trading interface. AI-generated fake documents, deepfakes, and cloned websites enabled the scheme. Exante, which does not operate in the U.S., confirmed the fraud and filed reports with multiple U.S. agencies.

Malicious Actors & Misuse · Date 2025-04-10 · Estimated Unknown

Reported developer
Openai, Generative Ai Developers
Reported deployer
Scammers Cloning Exante, Scammers
Reported affected
Individual Investors, Exante, Jpmorgan Chase
Source snapshot ↗

Personal voice assistants struggle with black voices, new study shows

A study found that voice recognition tools from Apple, Amazon, Google, IBM, and Microsoft disproportionately made errors when transcribing black speakers.

Discrimination and Toxicity · Date 2020-03-23 · Estimated true

Reported developer
Microsoft, Ibm, Google, Apple, Amazon
Reported deployer
Microsoft, Ibm, Google, Apple, Amazon
Reported affected
Black People
Source snapshot ↗

Reportedly Unsafe Deployment of Llama.cpp Reveals Interactive AI-Generated CSAM Roleplay Prompts

A study by UpGuard reports that misconfigured llama.cpp servers publicly exposed user prompts, including hundreds of interactive roleplay scenarios. Some prompts explicitly described fictional sexual abuse of children aged 7–12. While no real children were involved, the findings demonstrate how open-source LLMs can be exploited to generate AI-enabled child sexual abuse material (CSAM).

Privacy & Security · Date 2025-04-11 · Estimated Unknown

Reported developer
Meta, Users Of Llama.Cpp Servers
Reported deployer
Users Of Llama.Cpp Servers
Reported affected
General Public, Users Of Llama.Cpp Servers
Source snapshot ↗

Deepfake Job Applicant Allegedly Used AI Tools to Apply for Remote Role at U.S. Security Startup

Voice authentication startup Pindrop Security reported a job candidate used deepfake software and other AI tools in an attempted scam. This is part of a growing trend in international scammers using AI tools to apply for remote US-based jobs, sometimes successfully.

Malicious Actors & Misuse · Date 2025-04-08 · Estimated Unknown

Reported developer
Deepfake Technology Developers, Generative Ai Developers
Reported deployer
Unknown International Job Applicants, Scammers
Reported affected
Pindrop Security, Companies Hiring For Remote Positions
Source snapshot ↗

Deepfake Video Reportedly Depicts Zambian President Hakainde Hichilema Withdrawing from 2026 Election

A viral video circulating on social media in October 2023 falsely claimed that Zambian president Hakainde Hichilema announced he would not run in the 2026 elections. Fact-checkers and digital forensics experts confirmed the video was manipulated using basic AI tools. The footage originated from a TikTok account known for AI-generated content.

Malicious Actors & Misuse · Date 2023-10-12 · Estimated Unknown

Reported developer
Deepfake Technology Developers
Reported deployer
@Zambian_Ai Tiktok Account, Unknown Actors
Reported affected
Hakainde Hichilema, General Public Of Zambia, United Party For National Development (Upnd), Democratic Integrity
Source snapshot ↗

Purported AI-Cloned Voice Depicts Kenyan President William Ruto Speaking French in Viral Video

A purportedly AI-manipulated video widely shared on TikTok and Facebook presents Kenyan President William Ruto speaking in French during a speech in Tokyo on February 7, 2024. A reported reverse image search confirmed the original footage shows Ruto speaking in English to the Kenyan diaspora. The reportedly manipulated version used AI-generated French audio over the original video, making it a case of voice cloning rather than a full audiovisual deepfake.

Misinformation · Date 2024-02-19 · Estimated Unknown

Reported developer
Deepfake Technology Developers, Synthetic Audio Generation Technology Developers
Reported deployer
Unknown Actors Manipulating Footage Of William Ruto, Unknown Actors
Reported affected
William Ruto, General Public Of Kenya
Source snapshot ↗

Kenyan Journalist Jeff Koinange Depicted Endorsing Gambling App in Purported AI-Generated Deepfake

A viral video purportedly manipulated by AI shows Kenyan journalist Jeff Koinange endorsing a mobile gambling app. The video was reportedly created using AI deepfake techniques, uses footage from a legitimate news broadcast, and artificially alters Koinange's voice and mouth movements. Neither Koinange nor Citizen TV has promoted the app, and no legitimate source hosts the video.

Malicious Actors & Misuse · Date 2024-04-24 · Estimated Unknown

Reported developer
Deepfake Technology Developers
Reported deployer
Scammers Impersonating Jeff Koinange, Scammers
Reported affected
Jeff Koinange, General Public Of Kenya
Source snapshot ↗

Alleged Deepfake Video Depicts Former President of Kenya Uhuru Kenyatta Announcing 2027 Presidential Bid

A widely shared TikTok video purportedly manipulated by AI depicts former Kenyan president Uhuru Kenyatta announcing a 2027 presidential run. Reported digital forensics and visual inconsistencies confirm the video is an AI-generated deepfake, using archival footage from his presidency with synthesized audio. The clip spread rapidly online but has reportedly been debunked by fact-checkers. Kenyatta has made no such announcement, and such a move would violate Kenya's two-term constitutional limit.

Misinformation · Date 2024-03-18 · Estimated Unknown

Reported developer
Deepfake Technology Developers
Reported deployer
Unknown Actors, Information Manipulation Actors
Reported affected
Uhuru Kenyatta, General Public Of Kenya
Source snapshot ↗

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.

Discrimination and Toxicity · Date 2025-04-12 · Estimated Unknown

Reported developer
Anthropic, Openai, Deepseek Ai
Reported deployer
Anthropic, Openai, Deepseek Ai
Reported affected
Substack @Interruptingtea, General Public, Emotionally Vulnerable Individuals
Source snapshot ↗

Norwegian Supreme Court Receives Legal Filing with Fabricated Citations Allegedly Generated by AI Tool

In early 2025, Norway’s Supreme Court (Høyesterett) received a legal filing containing hallucinated legal citations and quotes, allegedly inserted by an attorney using an AI tool without verification. The fabricated sources were detected during standard review. While no sanctions were issued, the Court updated its guidance to lawyers, warning that AI use does not exempt them from professional responsibility.

Misinformation · Date 2025-04-10 · Estimated Unknown

Reported developer
Unspecified Generative Ai Developer
Reported deployer
Unnamed Norwegian Lawyer
Reported affected
Supreme Court Of Norway, Hoyesterett, Unnamed Clients Represented By Unnamed Norwegian Lawyer, Judicial Integrity
Source snapshot ↗

OpenAI's Operator Agent Reportedly Executed Unauthorized $31.43 Transaction Despite Safety Protocol

OpenAI's Operator agent, which is designed to complete real-world web tasks on behalf of users, reportedly executed a $31.43 grocery delivery purchase without user consent. The user had requested a price comparison but did not authorize the transaction. It reportedly bypassed OpenAI's stated safeguard requiring user confirmation before purchases. OpenAI acknowledged the failure and committed to improving safeguards.

AI system safety, failures, and limitations · Date 2025-02-07 · Estimated Unknown

Reported developer
Openai
Reported deployer
Openai
Reported affected
Users Of Operator, Geoffrey A. Fowler
Source snapshot ↗

Purported AI-Generated Audio Was Reportedly Used to Claim Recep Tayyip Erdoğan Expressed Support for Imran Khan in Pakistani Parliament

A reported viral video that was circulated in Pakistan featured purported AI-generated audio depicting Turkish President Recep Tayyip Erdoğan expressing support for jailed former Prime Minister Imran Khan during his 2025 visit. The video is described as having reused footage from 2020 and featured AI-generated audio that fabricated remarks Erdoğan never made. The synthetic voice was reportedly confirmed by forensic tools and audio analysis.

Misinformation · Date 2025-02-15 · Estimated Unknown

Reported developer
Deepfake Technology Developers, Synthetic Audio Generation Technology
Reported deployer
Unknown Actors
Reported affected
Recep Tayyip Erdogan, Pakistan Turkey Relations, Imran Khan, General Public Of Turkey, General Public Of Pakistan
Source snapshot ↗

Twitter’s Image Cropping Tool Allegedly Showed Gender and Racial Bias

Twitter's photo cropping algorithm was revealed by researchers to favor white and women faces in photos containing multiple faces, prompting the company to stop its use on mobile platform.

Discrimination and Toxicity · Date 2020-09-18 · Estimated true

Reported developer
Twitter
Reported deployer
Twitter
Reported affected
Twitter Non White Users, Twitter Non Male Users, X (Twitter) Users
Source snapshot ↗

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace