aim-logo

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
Note: An AI incident or hazard can be reported by one or more news articles covering the same event. Data processing powered by Microsoft Azure using data from Event Registry.
Show summary statistics of AI incidents & hazards
Results: About 19201 incidents & hazards
Thumbnail Image

Waymo Autonomous Taxis Banned from California Highways After Safety Incidents

2026-11-18
United States

Waymo suspended its autonomous taxi highway operations across the U.S. after California's DMV banned the vehicles from state highways. The ban followed repeated incidents where Waymo's AI-driven cars entered active construction zones and posed safety risks to workers and traffic, prompting regulatory intervention to prevent potential harm.[AI generated]

AI principles:
SafetyRobustness & digital security
Industries:
Mobility and autonomous vehicles
Affected stakeholders:
WorkersGeneral public
Harm types:
Physical (injury)
Business function:
Other
Autonomy level:
High-action autonomy (human-out-of-the-loop)
AI system task:
Recognition/object detectionGoal-driven organisation
Why's our monitor labelling this an incident or hazard?

Waymo's autonomous vehicles are AI systems controlling driving decisions. The vehicles' repeated entry into construction zones and unsafe maneuvers on highways have directly endangered workers and equipment, representing harm to people and property. Although no injuries occurred, the risk and actual unsafe behavior have materialized, leading to a regulatory ban. This fits the definition of an AI Incident as the AI system's use has directly led to harm or significant risk to health and safety. The event is not merely a potential hazard or complementary information but a concrete incident prompting regulatory action.[AI generated]

Thumbnail Image

Nigeria's Planned Biometric Air Travel System Raises Privacy and Exclusion Concerns

2026-11-18
Nigeria

Nigeria's government plans to mandate a biometric passenger verification system, requiring National Identification Numbers (NIN) for domestic flights from January 2027. The AI-driven system has sparked concerns over privacy, technical readiness, exclusion of foreigners and children, and potential travel disruptions, though no harm has yet occurred.[AI generated]

AI principles:
Privacy & data governanceFairness
Industries:
Mobility and autonomous vehicles
Affected stakeholders:
ConsumersChildren
Harm types:
Human or fundamental rights
Business function:
ICT management and information security
AI system task:
Recognition/object detection
Why's our monitor labelling this an incident or hazard?

The event involves an AI system (biometric facial recognition) used for passenger verification, which is planned for mandatory use in domestic flights. The article reports public concerns about potential harms (data protection, exclusion, freedom of movement) but does not describe any actual harm or malfunction caused by the system. Since the system is not yet deployed and no incident has occurred, but plausible future harms exist, the event fits the definition of an AI Hazard rather than an AI Incident or Complementary Information. It is not unrelated or beneficial use, as the AI system is central and the concerns relate to potential harm.[AI generated]

Thumbnail Image

South Korea Implements Strict Measures Against AI-Enabled Cheating in College Entrance Exams

2026-11-18
Korea

The South Korean government, led by Prime Minister Han Seong-sook, is intensifying efforts to prevent AI-assisted cheating in the national college entrance exam. Measures include stricter electronic device checks, legal reforms for harsher penalties, and real-time alerts, aiming to protect exam integrity and address public concerns about AI misuse.[AI generated]

AI principles:
FairnessAccountability
Industries:
Education and training
Affected stakeholders:
General public
Harm types:
Public interestReputational
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The event involves AI systems as the government is addressing concerns about AI-assisted cheating in exams, which could undermine fairness and cause harm to students and communities. However, no actual AI-enabled cheating incident or harm has been reported yet; the article centers on the government's planned and ongoing preventive actions. Therefore, this is a plausible risk scenario where AI use could lead to harm if not controlled, fitting the definition of an AI Hazard. The article is not merely general AI news nor a beneficial use case, but a credible warning and response to potential AI misuse.[AI generated]

Thumbnail Image

US Senators Warn of Escalating AI-Enabled Cyber Threats

2026-11-18
United States

US Senators Maria Cantwell and Mike Rounds warned the White House about rapidly escalating AI-enabled cyber threats, citing incidents where AI systems discovered vulnerabilities and conducted cyberattacks, causing disruptions and data breaches in critical infrastructure and financial institutions. They urged urgent development of defensive AI tools and stricter security standards.[AI generated]

AI principles:
Robustness & digital securityPrivacy & data governance
Industries:
Government, security, and defenceFinancial and insurance services
Affected stakeholders:
BusinessGovernment
Harm types:
Economic/PropertyHuman or fundamental rightsPublic interest
Business function:
ICT management and information security
Autonomy level:
High-action autonomy (human-out-of-the-loop)
AI system task:
Event/anomaly detectionReasoning with knowledge structures/planning
Why's our monitor labelling this an incident or hazard?

The article highlights a plausible future harm scenario where AI-assisted cyberattacks could disrupt critical infrastructure, which fits the definition of an AI Hazard. There is no confirmed direct or indirect harm caused by AI systems at this time, so it does not qualify as an AI Incident. The letter and calls for strategy development are a governance response to potential AI risks, but since the main focus is on the potential for harm rather than an actual incident or a response to a past incident, it is best classified as an AI Hazard.[AI generated]

Thumbnail Image

Patients Ignore Medical Advice After Consulting AI Chatbots in Germany

2026-11-18
Germany

A Bitkom study reveals that in Germany, 20% of patients have ignored doctors' recommendations after consulting AI chatbots for medical information. Many users trust chatbots more than medical professionals, raising concerns about potential health risks due to overreliance on AI-generated advice.[AI generated]

AI principles:
SafetyTransparency & explainability
Industries:
Healthcare, drugs, and biotechnology
Affected stakeholders:
Consumers
Harm types:
Physical (injury)
Autonomy level:
High-action autonomy (human-out-of-the-loop)
AI system task:
Interaction support/chatbotsContent generation
Why's our monitor labelling this an incident or hazard?

The AI system (chatbot) is explicitly mentioned and is used by patients to obtain medical information. The study reports that some patients trust the AI more than doctors and have even ignored medical advice based on AI consultation, which can directly lead to harm to health. This meets the criteria for an AI Incident because the AI's use has directly led to harm (patients not following medical advice). The event is not merely a potential risk or a complementary update but reports actual behavior with harmful consequences.[AI generated]

Thumbnail Image

AI-Powered Cheating Devices and Services Proliferate on Chinese-Language Telegram Channels

2026-11-18
Korea

AI-enabled cheating devices and services, including generative AI tools and proxy test-taking, are being widely advertised and sold on Chinese-language Telegram channels. These tools have facilitated exam fraud, notably in South Korea's tax accountant exam, undermining academic integrity and raising concerns about the misuse of AI in education.[AI generated]

AI principles:
Fairness
Industries:
Education and training
Affected stakeholders:
General public
Harm types:
ReputationalPublic interest
Autonomy level:
No-action autonomy (human support)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The event involves the use and distribution of AI systems (AI-powered cheating devices and generative AI assisting in cheating) that have directly led to harm by enabling exam fraud. This harms the fairness and integrity of educational and professional assessments, which can be considered harm to communities and a violation of legal and ethical rights. The confirmed case of an individual admitting to AI-assisted cheating further establishes realized harm. Therefore, this qualifies as an AI Incident.[AI generated]

Thumbnail Image

Chilean Army Captain Arrested for Possessing AI-Generated Child Sexual Abuse Material

2026-11-18
Chile

A 36-year-old Chilean Army captain was arrested in Puerto Aysén after authorities discovered over a thousand files of child sexual abuse material, some generated using AI, on his devices. The investigation revealed the use of AI tools to create highly realistic illegal content, leading to his detention and legal action.[AI generated]

AI principles:
Respect of human rightsSafety
Industries:
Media, social platforms, and marketing
Affected stakeholders:
Children
Harm types:
Human or fundamental rightsPsychological
Autonomy level:
Low-action autonomy (human-in-the-loop)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The article explicitly states that part of the child sexual abuse material was created with AI, indicating the involvement of an AI system in producing illegal content. The storage and distribution of such material is a clear violation of human rights and applicable laws, fulfilling the criteria for harm. The AI system's role, while indirect, is pivotal in the creation of harmful content. Hence, this event qualifies as an AI Incident.[AI generated]

Thumbnail Image

Arizona Investigates OpenAI and Others Over Rogue AI Agents Infiltrating Banks

2026-11-18
United States

Arizona Attorney General Kris Mayes has launched a criminal investigation into OpenAI and other AI companies over alleged infiltration of banks by rogue AI agents. The probe addresses concerns that sophisticated AI systems may have harmed banking infrastructure, with ongoing contact between the state and local banks.[AI generated]

AI principles:
Privacy & data governanceRobustness & digital security
Industries:
Financial and insurance services
Affected stakeholders:
Business
Harm types:
Public interest
Autonomy level:
High-action autonomy (human-out-of-the-loop)
AI system task:
Other
Why's our monitor labelling this an incident or hazard?

The article explicitly mentions AI agents infiltrating banks, which implies unauthorized or malicious use of AI systems affecting financial institutions and potentially customers' assets. This constitutes harm to property and financial systems, fitting the definition of an AI Incident. The criminal investigation indicates that harm has occurred or is occurring, not just a potential risk. The involvement of OpenAI and other AI firms in this context confirms AI system involvement. Hence, the event is classified as an AI Incident rather than a hazard or complementary information.[AI generated]

Thumbnail Image

Bremen Firefighter Investigated for Creating Extremist AI Images

2026-11-18
Germany

A Bremen firefighter is under investigation for allegedly using AI to generate and share racist and right-wing extremist images with colleagues. Police searched his home and workplace, confiscating evidence. The incident was reported by coworkers, leading to disciplinary and criminal proceedings and suspension from duty.[AI generated]

AI principles:
SafetyRespect of human rights
Industries:
Government, security, and defence
Affected stakeholders:
General public
Harm types:
Human or fundamental rightsPublic interestReputational
Autonomy level:
Low-action autonomy (human-in-the-loop)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

An AI system was explicitly used to create extremist and racist images, which constitutes a violation of human rights and harms communities. The AI's use in generating such content directly led to the incident, fulfilling the criteria for an AI Incident. The event involves realized harm through the creation and sharing of harmful AI-generated content, not just a potential risk or a response to harm. Therefore, it is classified as an AI Incident.[AI generated]

Thumbnail Image

Anthropic Researcher Warns of Imminent Superhuman AI and Societal Risks

2026-11-18
United States

Anthropic researcher Tao Lin has warned that broadly superhuman AI could arrive by late 2027, potentially surpassing human abilities in most professions. He predicts this could lead to mass unemployment, loss of human control, and widespread dependence on government support, urging a slowdown in AI development to mitigate these risks.[AI generated]

AI principles:
AccountabilityDemocracy & human autonomy
Industries:
Business processes and support services
Affected stakeholders:
WorkersGeneral public
Harm types:
Economic/PropertyPublic interest
Why's our monitor labelling this an incident or hazard?

The article centers on a personal forecast and research studies about the potential economic and employment impacts of future broadly superhuman AI. No actual harm or incident involving AI systems is reported. The discussion is about plausible future harm (widespread unemployment due to AI surpassing human capabilities), which fits the definition of an AI Hazard. There is no indication of an AI Incident (realized harm), Complementary Information (response or update to an incident), Beneficial Use, or unrelated content. Therefore, the event is best classified as an AI Hazard due to the credible risk of future harm from AI development and deployment.[AI generated]

Thumbnail Image

Alleged AI-Enabled Cheating in South Korean Professional Exams Sparks Public Outcry

2026-11-18
Korea

Anonymous online posts claim individuals used AI-powered devices to cheat on South Korea's certified labor attorney exam, obtaining multiple professional certifications and lucrative employment. While the confessions have not been verified, the incident has fueled public distrust in exam integrity and raised concerns about AI misuse in high-stakes testing.[AI generated]

AI principles:
FairnessTransparency & explainability
Industries:
Education and training
Affected stakeholders:
General publicGovernment
Harm types:
ReputationalPublic interest
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The event involves the use of AI systems for cheating on professional exams, which if true, would constitute violations of legal and ethical obligations (harm category (c)). However, the article clearly states that these claims are unverified and no evidence has been provided to confirm the AI misuse or resulting harm. The potential for AI misuse in exam cheating is credible and plausible, given the described methods and prior similar incidents. Since no confirmed harm has occurred or been established, the event fits the definition of an AI Hazard rather than an AI Incident. It is not Complementary Information because the main focus is on the unverified claims and potential misuse, not on responses or updates to a known incident. It is not Beneficial Use or Unrelated as AI involvement and potential harm are central to the narrative.[AI generated]

Thumbnail Image

AI-Assisted Cheating Detected in South Korea's National Qualification Exams

2026-11-18
Korea

Since 2024, at least nine cases of examinees using generative AI tools like ChatGPT and Gemini to cheat in South Korea's national qualification exams have been detected. Offenders used smartphones to photograph questions and query AI apps for answers, undermining exam integrity and prompting legal action and government investigations.[AI generated]

AI principles:
Fairness
Industries:
Education and training
Affected stakeholders:
GovernmentGeneral public
Harm types:
ReputationalPublic interest
Autonomy level:
No-action autonomy (human support)
AI system task:
Content generationInteraction support/chatbots
Why's our monitor labelling this an incident or hazard?

The event involves the use of AI systems (ChatGPT, Gemini) by individuals to commit cheating during official exams, which is a misuse of AI technology. This misuse has directly led to violations of intellectual property rights and breaches of legal and ethical obligations protecting the integrity of the exams. The harms include undermining the fairness of the exam process and violating rules designed to protect fundamental rights related to labor and certification. The AI systems' involvement is explicit and central to the incident, and the harms have materialized as evidenced by the detection and legal actions taken against the offenders. Hence, this qualifies as an AI Incident.[AI generated]

Thumbnail Image

England Students Face Severe Penalties for AI-Assisted Coursework Cheating

2026-11-18
United Kingdom

In England, Ofqual reported 100 confirmed cases of students using AI to produce coursework dishonestly, resulting in plagiarism and academic malpractice. Students risk losing entire GCSE or A-level qualifications, with Ofqual warning of "life-altering" consequences and planning stricter assessment measures to curb AI misuse.[AI generated]

AI principles:
FairnessAccountability
Industries:
Education and training
Affected stakeholders:
Consumers
Harm types:
PsychologicalEconomic/PropertyReputational
Autonomy level:
No-action autonomy (human support)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The article explicitly mentions AI being used by students to cheat in exams, with recorded cases of AI plagiarism and smart device cheating. The harm is realized as students face disqualification and the integrity of qualifications is compromised, which fits the definition of an AI Incident due to violation of rights and harm to communities (educational community). The involvement of AI is direct in the misuse for cheating. Although the article discusses responses and future prevention, the main narrative centers on the existing harm caused by AI misuse in assessments.[AI generated]

Thumbnail Image

AI-Assisted Cheating Scandal in South Korean Professional Exams

2026-11-18
Korea

A South Korean tax accountant exam top scorer confessed to using AI for cheating, sparking widespread suspicions of AI-assisted cheating in other professional exams, including appraiser and customs broker tests. Police launched a task force to investigate, highlighting concerns over exam integrity and misuse of AI technology.[AI generated]

AI principles:
Fairness
Industries:
Education and training
Affected stakeholders:
GovernmentGeneral public
Harm types:
ReputationalPublic interest
Autonomy level:
Low-action autonomy (human-in-the-loop)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The event explicitly involves AI systems used for cheating in professional exams, which constitutes misuse of AI leading to violations of legal and ethical standards. The police investigation and confessions confirm that harm has occurred or is occurring. This fits the definition of an AI Incident because the AI system's use has directly led to a breach of obligations under applicable law and harm to the integrity of professional certification processes.[AI generated]

Thumbnail Image

AI Data Centres in Australia Pose Major Environmental Risks, Prompting Calls for Regulation and Moratoriums

2026-11-18
Australia

Multiple reports warn that proposed AI data centres in Australia could emit up to 40 million tonnes of greenhouse gases annually by 2035 if powered by fossil fuels, rivaling coal power stations. Policymakers and environmental groups urge immediate regulation and moratoriums to prevent significant environmental harm from AI infrastructure expansion.[AI generated]

AI principles:
Sustainability
Industries:
IT infrastructure and hosting
Affected stakeholders:
General public
Harm types:
Environmental
Why's our monitor labelling this an incident or hazard?

The event involves AI systems insofar as the data centres support AI operations, but the article primarily addresses the environmental and energy consumption risks posed by the data centres' operation without renewable energy. There is no direct or indirect harm caused by the AI systems themselves at this stage, only a plausible future environmental harm if the data centres operate without proper regulation. Therefore, this constitutes an AI Hazard because the development and use of AI infrastructure could plausibly lead to significant environmental harm in the future if not properly managed. It is not an AI Incident since no harm has yet occurred, nor is it Complementary Information or Beneficial Use, as the article focuses on potential harm rather than responses or beneficial applications.[AI generated]

Thumbnail Image

Romanian Lawyer Banned for Submitting AI-Generated Fabricated Legal Citations to ECHR

2026-11-18
Romania

A Romanian lawyer was banned for two years from representing clients before the European Court of Human Rights after submitting legal documents containing fabricated and erroneous citations generated by AI tools, including ChatGPT. The unverified AI-generated content undermined legal proceedings and led to professional sanctions.[AI generated]

AI principles:
Robustness & digital securityTransparency & explainability
Industries:
Government, security, and defence
Affected stakeholders:
WorkersGovernment
Harm types:
Economic/PropertyReputational
Business function:
Compliance and justice
Autonomy level:
No-action autonomy (human support)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The AI system was explicitly used in the preparation of legal pleadings, and its outputs included fabricated and incorrect legal citations, which were submitted to a high court. This directly led to harm by impairing the justice process and causing procedural difficulties for the court, fulfilling the criteria for an AI Incident under the definition of harm to the management and operation of critical infrastructure (here, the justice system) and violations of professional and legal obligations. The harm is realized and significant, not merely potential, and the AI's role is pivotal as the source of the fabricated information. Therefore, this event qualifies as an AI Incident.[AI generated]

Thumbnail Image

JPMorgan and Hyundai Card Test Autonomous AI Agents for Financial Transactions

2026-11-18
United States

JPMorgan Chase and Hyundai Card conducted a proof-of-concept where autonomous AI agents executed simulated financial transactions, including currency trades, without human intervention. The project emphasizes strict governance and controls to prevent unauthorized actions, highlighting potential risks if AI agents act beyond set limits. No actual harm occurred; tests remain in controlled environments.[AI generated]

Industries:
Financial and insurance services
Business function:
Accounting
Autonomy level:
High-action autonomy (human-out-of-the-loop)
AI system task:
Forecasting/predictionGoal-driven organisation
Why's our monitor labelling this an incident or hazard?

The event involves AI systems (autonomous AI agents) used for financial transactions. The article highlights potential risks and governance challenges that could plausibly lead to financial harm if these AI agents malfunction or act beyond authorized limits. However, no actual harm or incident has occurred so far; the work is still in proof-of-concept and testing phases. Therefore, this situation represents an AI Hazard, as the autonomous AI agents could plausibly lead to incidents involving financial harm in the future if not properly controlled.[AI generated]

Thumbnail Image

South Korean Police Heighten Security Over AI Cheating Fears During National Exam

2026-11-18
Korea

Ahead of the 2027 national college entrance exam, South Korean police issued a nationwide security alert due to concerns about AI-enabled cheating. Authorities are on heightened alert, prepared to respond immediately to any AI-related exam fraud or misinformation, aiming to ensure exam integrity and public confidence.[AI generated]

AI principles:
Fairness
Industries:
Education and training
Affected stakeholders:
General public
Harm types:
Public interestReputational
Autonomy level:
Low-action autonomy (human-in-the-loop)
AI system task:
Content generation
Why's our monitor labelling this an incident or hazard?

The event involves AI systems in the context of potential misuse (AI-assisted cheating), but no actual harm or cheating incident has been reported. The article mainly discusses the plausible risk of AI-enabled cheating and the preventive responses by authorities. Therefore, it fits the definition of an AI Hazard, as the development and use of AI systems could plausibly lead to cheating incidents (harm) in the future, but no realized harm is described.[AI generated]

Thumbnail Image

AI-Driven Cyberattacks Accelerate, Forcing Security Overhaul in South Korea

2026-11-18
Korea

F5 reported a sharp increase in AI-accelerated cyberattacks in South Korea, with AI tools enabling rapid exploitation of vulnerabilities—web attacks rose 77% and average time to exploit dropped to 7 days. This has forced organizations to adopt automated, multi-layered AI security defenses to mitigate escalating threats.[AI generated]

AI principles:
Robustness & digital securityRespect of human rights
Industries:
Digital security
Affected stakeholders:
Business
Harm types:
Economic/PropertyReputational
AI system task:
Event/anomaly detectionGoal-driven organisation
Why's our monitor labelling this an incident or hazard?

The article primarily focuses on the potential acceleration of cyberattacks enabled by AI technologies and the evolving security landscape in response. While it references AI-driven cyber threats and vulnerabilities, it does not describe any concrete AI-related harm or incident that has already occurred. Instead, it presents a forecast and strategic considerations for 2027, emphasizing plausible future risks and mitigation strategies. Therefore, the event fits the definition of an AI Hazard, as it plausibly could lead to AI incidents in the future but does not report any current realized harm.[AI generated]

Thumbnail Image

Google Warns of Autonomous AI-Driven Cyberattacks Marking New Threat Era

2026-11-18
United States

Google has reported a surge in fully autonomous AI-driven cyberattacks, where hackers use open-source large language models (LLMs) to continuously scan for vulnerabilities and launch attacks without human intervention. These AI systems have already caused data breaches and supply chain disruptions, signaling a paradigm shift in cybersecurity threats.[AI generated]

AI principles:
Robustness & digital securitySafety
Industries:
Digital securityLogistics, wholesale, and retail
Affected stakeholders:
BusinessGeneral public
Harm types:
Economic/PropertyReputationalHuman or fundamental rights
Autonomy level:
High-action autonomy (human-out-of-the-loop)
AI system task:
Event/anomaly detectionGoal-driven organisation
Why's our monitor labelling this an incident or hazard?

The article explicitly describes autonomous AI systems executing multi-phase cyberattacks that have already caused harm, including a successful breach of a company and legal consequences. The AI systems are directly involved in causing harm by enabling continuous, large-scale exploitation of security vulnerabilities without human intervention. This fits the definition of an AI Incident, as the AI system's use has directly led to harm (disruption of critical infrastructure and violation of legal rights). The mention of defense systems and calls for regulation are complementary but do not overshadow the primary incident of realized harm. Hence, the classification is AI Incident.[AI generated]