TikTok uses AI systems to recommend and curate content, which is alleged to have exposed children to harmful sexual content and addictive features, leading to harm. The lawsuit claims violations of child protection laws and deceptive practices regarding content risks. The AI system's role in enabling addictive behavior and exposure to harmful content constitutes indirect harm to children, fitting the definition of an AI Incident. The event is not merely a legal or policy update but centers on harm caused by AI system use.[AI generated]
AIM: AI Incidents and Hazards Monitor
Automated monitor of incidents and hazards from public sources (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
Florida Sues TikTok Over AI-Driven Harmful Content and Addiction in Minors
Florida's Attorney General has sued TikTok, alleging its AI-powered recommendation system exposes children under 14 to violent, sexual, and addictive content, violating state child safety laws. The lawsuit claims TikTok's AI design deceives parents, enables underage access, and prioritizes profit over children's safety, seeking damages and compliance changes.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
Naver Cloud Unveils Lightweight AI Model for Military Applications in South Korea
Naver Cloud unveiled the HyperCLOVA X SEED 4B, a lightweight, multimodal AI model optimized for military use in South Korea. Designed for real-time analysis in edge environments like drones and tactical vehicles, its deployment in defense contexts poses potential future risks, though no harm has yet occurred.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves the development and planned deployment of an AI system tailored for military use, which inherently carries potential risks due to the high autonomy and critical nature of defense applications. However, the article does not describe any realized harm or incidents resulting from the AI system's use or malfunction. Instead, it focuses on the AI system's capabilities, optimization, and strategic roadmap for future deployment. Therefore, this event represents a plausible future risk scenario (AI Hazard) given the nature of military AI systems and their potential for harm if misused or malfunctioning, but no actual harm has been reported yet.[AI generated]
AI-Driven Deep Brain Stimulation Reduces Falls in Parkinson's Patients
Researchers in Switzerland and the US have developed AI-powered adaptive deep brain stimulation systems that interpret neural activity in real time to adjust stimulation for Parkinson's patients. These systems significantly improve walking ability and reduce falls, demonstrating direct health benefits and harm reduction through AI intervention.[AI generated]
Industries:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves an AI system embedded in an implanted neurostimulator that autonomously interprets neural signals and adjusts stimulation in real time. This AI system's use directly leads to a significant reduction in physical falls and improved gait in Parkinson's patients, which constitutes a direct health benefit and harm reduction. Since the AI system's use has directly led to improved health outcomes and reduced injury risk, this qualifies as an AI Incident under the definition of an event where AI use has directly led to injury or harm reduction (harm to health). The event is not merely a research announcement or product launch but reports realized health impacts from AI use in a clinical setting.[AI generated]
Serasa Accused of Massive Misuse of Biometric AI Data in Brazil
Unico, a digital identity and facial biometrics company, accuses Serasa Experian of unauthorized use of its AI-powered biometric technology, enabling millions of illicit biometric data queries on Brazilians. The incident, under civil and criminal investigation, led to a search and seizure operation in São Paulo and highlights large-scale privacy violations.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event explicitly involves AI systems (biometric facial recognition technology) and the unauthorized use of sensitive biometric data, which directly leads to violations of privacy and confidentiality rights. The involvement of AI in processing biometric data and the alleged misuse causing harm to individuals' rights fits the definition of an AI Incident. The legal actions and forensic findings confirm that harm has occurred, not just a potential risk, thus excluding classification as a hazard or complementary information.[AI generated]
Tesla Faces Scrutiny Over Misleading FSD Safety Data in Europe
Tesla presented self-published and potentially misleading safety statistics to European regulators in Sweden and the Netherlands to secure approval for its Full Self-Driving (FSD) AI system. Independent researchers and authorities questioned the validity of Tesla's claims, raising concerns about regulatory approval based on exaggerated safety data.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event explicitly involves an AI system (Tesla's FSD autonomous driving system) and concerns the use and presentation of data about its safety. The misleading statistics were presented to regulators, which could indirectly lead to harm if the system is approved and used under false safety assumptions. Since the article does not report actual harm occurring but highlights deceptive practices that could plausibly lead to harm, this qualifies as an AI Hazard rather than an AI Incident. The AI system's involvement is clear, and the potential for future harm is credible due to the risk of unsafe deployment based on false data.[AI generated]
Meta Tests Military-Grade Facial Recognition in Smart Glasses, Raising Privacy Concerns
Meta tested facial recognition and liveness detection software from Rank One Computing—a defense contractor with U.S. military and law enforcement clients—in its smart glasses. The AI features were embedded but never activated, and removed after discovery, highlighting significant privacy and surveillance risks if such technology is deployed in consumer devices.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves the development and potential use of an AI system (facial recognition and liveness detection) that is closely tied to surveillance technology used by law enforcement and military agencies. The article does not report any realized harm but highlights credible concerns about future misuse and privacy violations. Therefore, it fits the definition of an AI Hazard, as the AI system's use could plausibly lead to violations of human rights and privacy in the future. It is not an AI Incident because no harm has yet occurred, nor is it Complementary Information or Unrelated.[AI generated]
Greek TV Host Targeted by Harmful AI-Generated Fake Wedding Photos
Greek TV presenter Katerina Zarifi was targeted for the third time by AI-generated fake images depicting her as a bride, leading to widespread misinformation, emotional distress, and reputational harm. The incident caused confusion among her acquaintances and family, prompting her to pursue legal action against those responsible for creating and spreading the AI-manipulated content.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves AI systems generating fake images that misrepresent a person, leading to harm in the form of violation of personal rights and emotional distress. The AI-generated content is actively circulating and causing real harm, not just a potential risk. Therefore, this qualifies as an AI Incident due to the realized harm caused by the AI system's use in creating and spreading false personal images.[AI generated]
AI-Powered Brain-Computer Interface Restores Speech for Paralyzed Patient at Home
Researchers from the University of California, Davis, and collaborators developed an AI-driven brain-computer interface (BCI) that enabled a patient with severe paralysis to communicate independently from home. The system, using implanted electrodes and intelligent software, translated neural signals into text and cursor control, significantly improving the patient's autonomy and communication speed.[AI generated]
Industries:
Severity:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves an AI system (the BCI with implanted electrodes and intelligent software that translates neural activity into text and cursor control). The system's use has directly led to a positive health impact by enabling communication for a patient with severe paralysis, which qualifies as injury or harm to health being addressed. Although the article focuses on a positive outcome rather than harm, the framework includes harms to health, and this event is about mitigating such harm through AI. Since the event describes realized effects of the AI system on health and autonomy, it qualifies as an AI Incident (positive health impact through AI use). It is not a hazard (no plausible future harm), nor complementary information or unrelated.[AI generated]
AI-Powered Wireless Backdoor Detection System Deployed in Semiconductor Industry
Korean security firm Jison has signed contracts to supply AI-based wireless backdoor hacking detection systems to global semiconductor manufacturers. The system monitors server rooms for unauthorized wireless signals, addressing rising AI-driven hacking threats. Deployment aims to protect critical infrastructure and prevent potential data breaches, expanding from banks to manufacturing sectors.[AI generated]
Industries:
Severity:
Business function:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves the use of an AI system designed to detect wireless backdoor hacking attempts, which are a serious cybersecurity threat. While the article does not report a specific realized harm or incident, it clearly identifies the increasing AI-driven hacking threat and the deployment of AI-based detection systems as a response. The contracts and deployment of this system aim to prevent potential harms such as data breaches, intellectual property theft, or disruption of critical infrastructure. Since no actual harm has been reported yet but the AI system is used to mitigate a credible and significant cybersecurity risk, this qualifies as an AI Hazard, reflecting a plausible future harm scenario related to AI-powered hacking threats and their detection.[AI generated]
German Pension Insurance Warns of AI-Generated Phishing Emails
Criminals are using AI to craft convincing phishing emails mimicking the German Pension Insurance's official style, aiming to steal personal data from recipients. The organization has issued warnings, urging people not to follow instructions or click links in suspicious emails. The incident is centered in central Germany.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
An AI system is involved as the fraudsters use AI to generate convincing phishing emails. The event involves the use of AI in a malicious way to deceive people and obtain sensitive data, which constitutes a violation of personal rights and potentially leads to harm to individuals. Since the harm (fraud attempts targeting personal data) is occurring or actively attempted, this qualifies as an AI Incident. The AI system's use is in the malicious generation of phishing content that directly leads to harm or risk of harm to individuals' privacy and security.[AI generated]
U.S. Export Controls on Anthropic's AI Model Disrupt South Korea's Cybersecurity Collaboration
The U.S. government imposed export controls on Anthropic's advanced AI models, restricting access for foreign entities, including South Korean government and companies participating in the Glasswing cybersecurity project. This action threatens South Korea's AI-based cyber defense collaboration, raising concerns about national security and the need for independent AI capabilities.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly involves an AI system (Anthropic's 'Mitos' AI model) and its use by South Korean companies and institutions. The U.S. government's export control restricts access to this AI system, which could plausibly lead to harm by disrupting AI development, collaboration, and national AI sovereignty. No actual injury, rights violation, or direct harm has been reported yet, but the restriction poses a credible risk of harm in the future. Hence, it fits the definition of an AI Hazard rather than an AI Incident or Complementary Information.[AI generated]
Tesla Accused of Misleading Regulators on FSD Safety; AI Driver Monitoring Risks Highlighted
Tesla allegedly provided misleading safety data about its Full Self-Driving (FSD) AI system to European regulators, raising concerns over road safety and regulatory compliance. Separately, misuse of Tesla's AI driver monitoring in Taiwan, where drivers trick the system, exposes further risks of AI-assisted driving. No direct harm reported yet, but potential for incidents is high.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly involves an AI system (Tesla's supervised autonomous driving with AI-based driver monitoring). The misuse described (using props to trick the monitoring system) directly undermines the AI system's safety function, creating a credible risk of accidents and harm. However, the article does not report any actual accident or injury resulting from this misuse, only the potential for such harm. Hence, it fits the definition of an AI Hazard rather than an AI Incident. The discussion of legal and regulatory challenges further supports the recognition of plausible future harm.[AI generated]
AI-Generated Deepfake Scam Causes Major Financial Losses in France
In Pont-à-Mousson, France, an elderly woman lost up to €25,000 after being deceived by an AI-generated deepfake video featuring Emmanuel Macron and Bernard Arnault promoting a fraudulent investment platform. The scam, enabled by convincing AI-created content, led to significant financial harm and highlights the growing misuse of deepfakes in online fraud.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves the use of an AI system to create a deepfake video that was instrumental in perpetrating a financial scam. The victim suffered direct financial harm due to reliance on the AI-generated content. This meets the criteria for an AI Incident because the AI system's use directly led to harm to a person. The harm is realized and significant, not merely potential, and the AI system's role is pivotal in enabling the scam.[AI generated]
China Develops AI-Powered Robotic 'Machine Wolves' for Potential Taiwan Conflict
China is developing AI-enabled quadruped robots, dubbed 'machine wolves,' for military use in a potential Taiwan conflict. These robots, equipped with sensors and possible weapons, could be deployed for reconnaissance and logistics, potentially lowering the political cost of early casualties and altering conflict dynamics. Experts urge Taiwan to adopt multi-layered defenses.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly describes AI systems (robotic 'machine wolves' with sensors, AI computing, and autonomous capabilities) being developed and tested by the Chinese military for use in conflict scenarios. The use of these AI systems could plausibly lead to harm including injury or death in warfare, disruption of regional stability, and escalation of conflict. The article does not report any actual incident of harm caused by these AI systems but discusses their potential deployment and strategic impact. Therefore, it fits the definition of an AI Hazard, as it describes an event/circumstance where AI system development and intended use could plausibly lead to an AI Incident in the future. It is not Complementary Information because it is not an update or response to a past incident, nor is it unrelated as it clearly involves AI systems with military applications and potential harm.[AI generated]
AI Adoption Leads to Significant Job Losses Among Young Workers in South Korea
South Korea experienced its first decline in stable employment in 26 years, with 20- and 30-year-olds in information technology and professional sectors hit hardest. The reduction is attributed to both manufacturing downturns and AI-driven job displacement, causing economic harm and increased job insecurity for young workers.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves the use of AI systems in the workplace leading to a reduction in stable employment, which constitutes harm to communities and individuals through economic displacement and job loss. Although the harm is economic and social rather than physical, it fits within the framework's definition of harm to communities or significant articulated harm where AI's role is pivotal. Therefore, this qualifies as an AI Incident due to realized harm caused by AI's impact on employment.[AI generated]
AI Surveillance Cameras Prevent Youth Gang Violence in Phú Thọ
In Phú Thọ, Vietnam, AI-powered surveillance cameras enabled police to quickly identify and apprehend 18 teenagers preparing for a violent public disturbance. The AI system's facial recognition capabilities allowed authorities to intervene within 30 minutes, preventing potential harm to public safety and order.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
An AI system (camera AI for facial recognition) was explicitly mentioned and used by the police to identify individuals involved in a violent and unlawful gathering. The AI system's use directly led to the identification and subsequent legal processing of the youths, addressing harm to community safety and public order (harm to communities). Therefore, this qualifies as an AI Incident because the AI system's use directly contributed to mitigating harm related to public disorder and violence.[AI generated]
US Government Orders Shutdown of Anthropic's Advanced AI Models Over Security Concerns
The US government ordered Anthropic to globally disable access to its advanced AI models, Claude Fable 5 and Mythos 5, citing national security risks. The models, capable of identifying software vulnerabilities, were shut down shortly after release due to concerns they could be exploited, despite no reported actual harm.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves AI systems (Claude Fable 5 and Mythos 5) whose use is suspended due to a government order citing national security risks from a jailbreak technique. No actual harm has occurred yet, but the concern is that the AI models could be exploited to cause harm, such as security breaches. This fits the definition of an AI Hazard, as the development and use of these AI systems could plausibly lead to an AI Incident involving harm to critical infrastructure or security. The article does not describe any realized harm, only the potential risk and regulatory response, so it is not an AI Incident. It is not Complementary Information because the main focus is on the regulatory action and the risk, not on updates or responses to a past incident. Therefore, the classification is AI Hazard.[AI generated]
Hyundai Rotem Unveils AI-Based Multi-Layered Counter-Drone Defense at Eurosatory 2026
Hyundai Rotem showcased its AI-based unmanned turret and multi-layered counter-drone defense system at Eurosatory 2026 in Paris. The system autonomously detects, classifies, and responds to drone threats, combining electronic and physical countermeasures. While no harm has occurred, its military use poses plausible future risks.[AI generated]
AI principles:
Industries:
Severity:
Business function:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves an AI system (AI-based multi-layered defense system) under development and demonstration, with potential military applications. However, there is no indication that the system has caused any harm or malfunction, nor that any incident has occurred. The article presents the system as a future capability and a strategic product offering, implying potential future risks but no current harm. Therefore, this qualifies as an AI Hazard because the development and potential deployment of such AI-enabled military defense systems could plausibly lead to harms related to military conflict or misuse, but no incident has yet occurred.[AI generated]
AI-Enabled Scams Cause Surge in UK Financial Fraud Losses
Criminals in the UK have increasingly used AI to create sophisticated scams, including investment and authorised push payment fraud, resulting in over £576 million lost in 2025. AI has enabled large-scale, convincing communications, lowering barriers for fraudsters and causing significant financial and emotional harm to victims.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
The article explicitly states that criminals use AI to mimic voices and create fake profiles to scam victims, which has directly led to financial losses and emotional harm. This fits the definition of an AI Incident because the AI system's use in the fraud has directly caused harm to people. The harms include injury to people (emotional and financial harm) and harm to communities (widespread fraud). The involvement of AI is clear and central to the incident, and the harm is realized, not just potential. Therefore, this event qualifies as an AI Incident.[AI generated]
Family Trapped After Google Maps Error Leads Car Into Drain in Hapur
In Hapur, Uttar Pradesh, a family following Google Maps' navigation was led onto a hazardous route, causing their car to fall into a deep drain. The incident required a two-hour rescue operation to save all five family members. The AI system's inaccurate guidance directly contributed to the dangerous situation.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
Autonomy level:
AI system task:
Why's our monitor labelling this an incident or hazard?
Google Maps uses AI for route planning and navigation. The family's reliance on this AI system's guidance led them into a hazardous situation where their car fell into a drain. This is a clear example of harm indirectly caused by the use of an AI system, fulfilling the criteria for an AI Incident due to injury risk and disruption caused by the AI system's malfunction or error in routing.[AI generated]