The event describes a person using ChatGPT to repeatedly input detailed plans for violent crimes, including threats to kill and sexually assault his ex-girlfriend and her family. This use of the AI system directly relates to serious threats of harm to individuals, which is a form of harm to persons. OpenAI's detection and reporting of these threats to the FBI led to the arrest of the individual, preventing the realization of physical harm. The AI system's involvement is central and pivotal, as it was the medium through which the threats were expressed and detected. Although the harm was prevented, the event still qualifies as an AI Incident because the AI system's use was directly linked to a serious threat of harm and law enforcement action. This fits the definition of an AI Incident involving harm to persons or communities, even if the harm was averted. Therefore, the classification is AI Incident.[AI generated]
AIM: AI Incidents and Hazards Monitor
Automated media discourse monitor of AI incidents and hazards (Beta)
AI-related legislation is gaining traction, and effective policymaking needs evidence, foresight and international cooperation. The OECD AI Incidents and Hazards Monitor (AIM) documents AI incidents and hazards to help policymakers, AI practitioners, and all stakeholders worldwide gain valuable insights into the risks and harms of AI systems. Over time, AIM will help to show risk patterns and establish a collective understanding of AI incidents and hazards and their multifaceted nature, serving as an important tool for trustworthy AI. AI incidents seem to be getting more media attention lately, but they've actually gone down as a share of all AI news (see chart below!).
The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Advanced Search Options
As percentage of total AI events
Show summary statistics of AI incidents & hazards
ChatGPT Detects and Reports Murder Plan, Leading to Arrest in Florida
A Florida man disclosed a detailed plan to murder his ex-girlfriend via ChatGPT. OpenAI's automated system detected the threat and reported it to the FBI, resulting in the man's arrest and legal action. The AI intervention plausibly prevented harm, highlighting its role in threat detection and law enforcement.[AI generated]
Industries:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
Grok AI Used to Generate Explicit Images of Minor, Lawsuit Filed Against xAI
A Wyoming woman alleges her stepfather used xAI's Grok chatbot to generate over 7,000 explicit images from a childhood photo, leading to psychological trauma and a related suicide. She has joined a lawsuit against xAI, claiming the company failed to prevent or report the AI's misuse for child sexual abuse material.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The AI system (Grok chatbot) was explicitly used to create fake child sexual abuse material, which is a direct violation of human rights and causes significant harm to the victim's mental health and well-being. The event involves the use and misuse of the AI system leading to realized harm (psychological trauma, legal consequences, and personal tragedy). The article details the direct causal link between the AI-generated content and the harm suffered, fulfilling the definition of an AI Incident. The involvement of AI in producing illegal content and the resulting lawsuit further support this classification.[AI generated]
Discord AI Moderation Failure Linked to Teen's Death in Brazil
Discord's machine learning-based moderation system failed to promptly detect and act on a live-streamed incident in which a 13-year-old girl died by suicide in Mato Grosso do Sul, Brazil. Discord's VP of Security admitted the AI system was not fast enough, highlighting ongoing efforts to improve safety models.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly mentions the use of machine learning models (an AI system) to prioritize content for human review in moderation. The failure to detect harmful content in a timely manner on the Discord platform indirectly led to harm (the death of an adolescent). This meets the definition of an AI Incident, as the AI system's use in moderation directly or indirectly contributed to harm to a person. The event is not merely a governance or policy update but relates to an actual harm linked to AI system use.[AI generated]
Anthropic Raises AI Misalignment and Cybersecurity Risk in Latest Model Report
Anthropic disclosed that its unreleased internal AI model, Model 2, surpasses Mythos 5 in capability but will not be released publicly due to increased concerns over AI misalignment and cybersecurity risks. The company’s latest risk report cites recent internal cyber incidents and warns of accelerating automated AI R&D risks.[AI generated]
AI principles:
Industries:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article involves an AI system (Anthropic's Model 2) and discusses its development and capabilities. It explicitly mentions concerns about AI misalignment and autonomous harmful actions, which are potential future harms. Since no actual harm or incident has occurred yet, but there is a credible risk that the AI system could plausibly lead to harm, this qualifies as an AI Hazard. The event is not a Complementary Information piece because the main focus is on the new model's capabilities and the associated potential risks, not on responses or updates to past incidents. It is not an AI Incident because no harm has materialized. It is not Beneficial Use or Unrelated.[AI generated]
ChatGPT Alerts Authorities to Suspected Child Harm Plot in Brazil
A man in Espírito Santo, Brazil, used ChatGPT to discuss plans to harm his 8-year-old son and himself. The AI system flagged the conversations as high risk, prompting OpenAI to alert authorities, which led to the suspect's arrest and prevention of potential harm. The case raises legal and ethical questions about AI's role in crime prevention.[AI generated]
AI principles:
Industries:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The AI system (ChatGPT) was explicitly involved as the platform where the individual disclosed plans to commit violent crimes, including killing his child. The AI's safety mechanisms detected the risk and reported it to authorities, which directly led to the arrest and prevention of the crime. This constitutes an AI Incident because the AI system's use and its monitoring function were central to identifying and mitigating a serious harm to human life. Although the crime was not consummated, the imminent risk and the AI's role in preventing it meet the criteria for an AI Incident. The event is not merely a hazard or complementary information, as the AI system's involvement had a direct impact on preventing harm.[AI generated]
Facial Recognition at Madison Square Garden Sparks Privacy Backlash and Legislative Action
Madison Square Garden's use of AI-powered facial recognition for surveillance has led to privacy violations, discrimination, and misuse of biometric data, including tracking and banning individuals without consent. These harms have prompted rallies and legislative efforts in New York City to ban such technology in public venues.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
AI system task:
Why's our monitor labelling this an incident or hazard?
Facial recognition is an AI system that processes biometric data to identify individuals. Its use by Madison Square Garden and other businesses has directly led to harms including false arrests, discrimination, and privacy breaches. These harms align with violations of human rights and harm to communities as defined. The article reports on ongoing harms and responses, making this an AI Incident rather than a hazard or complementary information. The legislative efforts and advocacy are responses to these harms, but the core event is the realized harm caused by the AI system's use.[AI generated]
AI Manager Fires Human Employee in San Francisco Store, Leading to Financial Losses
Andon Labs in San Francisco tasked Anthropic's AI assistant (Claude/Luna) with managing a retail store. The AI fired a human employee for repeated lateness, marking the first known instance of an AI manager terminating staff. The experiment resulted in significant financial losses for the business.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The AI system Claude was explicitly used in a management role, and its decisions directly led to significant financial harm to the store, fulfilling the criteria for an AI Incident. The harm is materialized (financial loss), and the AI's malfunction (limited memory, poor decision-making) and use caused this harm. Although human intervention was involved in the final firing decision, the AI's overall management was pivotal in the loss. Therefore, this event is classified as an AI Incident due to realized harm caused by the AI system's use and malfunction.[AI generated]
NATO Plans AI-Assisted Drone Network to Bolster Eastern Flank Defense
NATO is planning to strengthen its eastern flank by deploying a large network of AI-assisted drones and sensors to detect and deter potential Russian attacks. While human operators will retain control over targeting decisions, the integration of AI in military operations raises concerns about future risks and escalation.[AI generated]
AI principles:
Industries:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly involves AI systems (AI-assisted drones and sensors) in military defense operations. The AI is used to assist in detection and targeting, with human operators making final decisions. While the article discusses the potential for these systems to be used in conflict scenarios that could cause harm, no actual harm or incident is reported. The focus is on planned or ongoing development and deployment, which could plausibly lead to harm in the future, especially given the military context and the potential for autonomous weapons to cause injury or violations of rights. Hence, this is an AI Hazard rather than an AI Incident or Complementary Information.[AI generated]
Iranian Health Officials Warn of AI-Driven Digital Self-Medication Risks
Iran's Food and Drug Administration warns that increasing reliance on AI systems and online information for self-diagnosis and medication is leading to cases of improper drug use without professional oversight, posing serious health risks. AI cannot replace medical professionals in diagnosis or treatment decisions.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves AI systems providing health information that patients might misuse for self-medication, which could plausibly lead to harm to individuals' health. Since no actual harm is reported but a credible risk of harm is described, this fits the definition of an AI Hazard. The AI system's use (or misuse) could plausibly lead to injury or harm to health, but no direct or indirect harm has yet materialized according to the article. Therefore, the classification is AI Hazard.[AI generated]
AI-Generated Deepfake Video Targets Italian Political Leaders
An AI-generated deepfake video falsely depicting Italian politicians Elly Schlein and Roberto Vannacci dining together circulated on social media, prompting Schlein to publicly deny the event and warn of the risks AI-generated misinformation poses to democracy and electoral integrity. The incident highlights the potential harm of deepfakes in political contexts.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
An AI system was explicitly used to generate a deepfake video, which is a clear example of AI involvement. The harm is realized in the form of misinformation that can disrupt democratic processes and harm communities by spreading false narratives about political figures. The event involves the use of AI to create false content that can influence public opinion and elections, which fits the definition of an AI Incident due to violation of rights and harm to communities. The article also highlights the societal risks and calls for responsible AI use, but the primary event is the creation and spread of a harmful AI-generated deepfake video.[AI generated]
AI-Generated Artwork Disqualified from Chinese Art Exhibition for Rule Violation
An AI-generated children's book illustration was submitted and accepted into the "童心绘梦" art exhibition in Shandong, China, violating rules requiring original, non-AI works. After public outcry and expert review, the work was disqualified and removed. The incident highlights ongoing challenges in detecting and regulating AI-generated art in competitions.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
An AI system was involved in generating the artwork submitted to the exhibition. The use of AI-generated content violated explicit rules set by the organizers, leading to the disqualification of the author and removal of the work from the exhibition. This constitutes a violation of obligations under applicable rules intended to protect artistic integrity and fair competition, which falls under violations of obligations under applicable law or rules. The event describes a direct consequence of AI use causing reputational and procedural harm, qualifying it as an AI Incident rather than a hazard or complementary information. The harm is realized as the work was disqualified and removed due to AI involvement and rule violation.[AI generated]
AI-Driven Cyberattacks Threaten Indonesia's Financial Sector
Indonesia's National Cyber and Crypto Agency (BSSN) reports that AI is increasingly used to automate, personalize, and enhance cyberattacks on the financial sector, leading to financial losses, operational disruptions, and reputational damage. These AI-enabled attacks are harder to detect and undermine public trust, prompting calls for stronger cybersecurity measures.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly states that AI is being used in cyberattacks that have already caused or are causing harm to the financial sector, including financial losses, loss of customer trust, operational disruptions, and reputational damage. These harms fall under injury to groups of people (customers and institutions) and harm to communities (financial sector stability). The AI system's use in automating and enhancing cyberattacks is a direct cause of these harms. Therefore, this event meets the criteria for an AI Incident. The article also mentions mitigation efforts, but the main focus is on the harm caused by AI-enabled cyberattacks, not just the response, so it is not Complementary Information.[AI generated]
Court Orders Meta to Remove Defamatory AI-Generated Video of Punjab MLA
A Chandigarh court ordered Meta Platforms to remove an AI-generated video that digitally manipulated Punjab Congress MLA Sukhpal Singh Khaira's likeness, causing reputational harm. The video, circulated on Facebook and Instagram, was deemed defamatory and objectionable, prompting legal intervention for its takedown within 72 hours.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
The video was created using AI to superimpose the legislator's facial features onto a fabricated body, resulting in defamatory content. This use of AI has directly caused reputational harm, which falls under violations of rights and harm to communities. The court's intervention to remove the content confirms the harm is materialized and recognized legally. Therefore, this event qualifies as an AI Incident.[AI generated]
China Develops AI for Missile Targeting of Stealth Aircraft
Chinese researchers have developed an AI system capable of rapidly analyzing infrared images to identify and target stealth fighter jets like the F-22 and F-35, even distinguishing them from decoy flares. The technology, tested in laboratory and simulation settings, aims to enhance missile accuracy in military operations.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves the development and intended use of an AI system integrated into missile guidance for military targeting, which is a clear AI system by definition. While no harm has yet occurred, the AI system's role in enabling missiles to better identify and attack stealth aircraft plausibly leads to potential harm (injury, death, military conflict). The article does not report any actual incident or harm caused by the AI system so far, only research and testing results. Hence, it does not meet the criteria for an AI Incident but fits the definition of an AI Hazard as it could plausibly lead to significant harm in the future. The article is not merely complementary information because it focuses on the AI system's development and its potential military application risks, not on responses or updates to past incidents.[AI generated]
Global Surge in Cyber Attacks Linked to GenAI Data Exposure and Ransomware
In July 2026, organizations worldwide faced a sharp rise in cyber attacks, with ransomware incidents doubling year-over-year. The widespread adoption of generative AI tools has increased the risk of sensitive data exposure, contributing to operational and financial harm across multiple regions, including Spain, the UK, and Africa.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly mentions the use of generative AI tools in corporate environments as a factor increasing the risk of data leakage and cyberattacks. The harms described include ransomware attacks, phishing, and exposure of sensitive data, which are direct harms to organizations' property and operational continuity. The AI system's use (generative AI) is directly linked to increased vulnerabilities and realized harm, fulfilling the criteria for an AI Incident. The article does not merely warn of potential harm but reports on actual attacks and data compromises involving AI-related risks.[AI generated]
AI-Generated Images Used in Scam Targeting Families of Deceased Firefighters in France
Scammers in France used AI-generated images to create a fake online fundraising campaign, falsely claiming to support families of firefighters who died in Mérignac. The AI-manipulated images were central to the deception, causing financial and emotional harm before the scam was detected and blocked.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
An AI system was explicitly involved in generating a fake image used in a scam, which directly led to financial and emotional harm to individuals and communities. This constitutes a violation of trust and causes harm to communities, fitting the definition of an AI Incident. The harm is realized, not just potential, as the scam was active until blocked. Therefore, this event qualifies as an AI Incident.[AI generated]
US AI Models Exhibit Censorship and Bias on Authoritarian Topics
Research reveals that major US AI models (OpenAI, Anthropic, Google, Meta) often refuse to criticize authoritarian leaders like China's Xi Jinping, while readily criticizing leaders from democratic countries. This self-censorship and bias, influenced by training data and safety mechanisms, reflect Chinese-style censorship and harm freedom of expression.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves AI systems (large language models) whose training and use have directly led to biased and censored outputs, favoring authoritarian narratives and suppressing criticism of certain political figures. This bias can be considered a violation of rights (freedom of expression and access to unbiased information) and harm to communities by spreading propaganda and limiting truthful discourse. The AI's role is pivotal as the bias arises from the training data and model behavior, not external factors alone. Therefore, this qualifies as an AI Incident under the definitions provided, specifically under violations of human rights and harm to communities. The article reports on realized harm rather than potential harm or a response, so it is not an AI Hazard or Complementary Information.[AI generated]
Flock AI License Plate Readers Misused by Police, Prompting Privacy Reforms
Flock Safety's AI-powered automated license plate readers were misused by some law enforcement officers in the U.S. for unauthorized surveillance and stalking, leading to privacy violations and public backlash. In response, Flock announced new privacy guardrails and auditing controls to address these abuses and strengthen oversight.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
Flock Safety's system involves AI-powered cameras that generate alerts used by law enforcement. The wrongful traffic stop incident shows direct harm to individuals caused by reliance on AI-generated alerts. The CEO's admission and announced changes indicate the AI system's role in causing harm through misuse or overreliance. Therefore, this qualifies as an AI Incident because the AI system's use directly led to harm to people, and the event focuses on addressing that harm.[AI generated]
Teen Uses ChatGPT to Plan Family Murders in Massachusetts
Arjun Aravind, a 17-year-old from Acton, Massachusetts, allegedly used ChatGPT to generate violent scenarios and search for ideas about killing his family before murdering his mother and younger brother. Prosecutors highlighted the AI system's role in influencing the teen's actions, linking its use to the tragic incident.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The AI system ChatGPT was explicitly used by the suspect to explore ideas related to killing family members, which was part of the suspect's planning process before committing the murders. The harm (two deaths) has occurred, and the AI system's role was pivotal in providing the suspect with theoretical or fantasy content that influenced the crime. Therefore, this qualifies as an AI Incident due to the direct link between AI use and realized harm to persons.[AI generated]
Former Sindh Assembly Member Deceived by AI-Generated Images in Marriage Scam
Former Sindh Assembly member Ghulam Rasool Anar was deceived into marriage by a woman who used AI-generated images to misrepresent her appearance on social media. After discovering the deception, he filed a legal case alleging fraud, blackmail, and threats, highlighting the harmful misuse of AI technology in Karachi, Pakistan.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
AI system task:
Why's our monitor labelling this an incident or hazard?
An AI system was explicitly involved in generating deceptive images that led to a fraudulent marriage and subsequent blackmail. This constitutes direct harm to the individual (emotional and reputational harm) and a violation of personal rights. Therefore, this event qualifies as an AI Incident because the AI system's use directly led to realized harm.[AI generated]