The article explicitly details AI systems (facial recognition, data integration, biometric surveillance algorithms) used by the Chinese government to monitor citizens and suppress dissent, which constitutes violations of human rights and harm to communities. The use of these AI systems is ongoing and has already led to harms such as repression of Uyghurs and surveillance beyond borders. The AI systems' development and use are directly linked to these harms, fulfilling the criteria for an AI Incident rather than a hazard or complementary information.[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
DeepSeek Security Risks Raise CCP Surveillance Concerns
Taiwan's Digital Development Minister Huang Yannan warned that the Chinese AI system DeepSeek poses serious data security risks, potentially allowing CCP surveillance. DeepSeek failed cybersecurity tests and is banned from government use in Taiwan, spurring plans for alternative AI services in the future.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Why's our monitor labelling this an incident or hazard?
DBS Bank's AI Shift Prompts Major Workforce Restructuring
DBS Bank in Singapore announced plans to replace 4,000 contract workers with AI systems over the next three years. CEO Piyush Gupta stated that while automation will reduce the workforce by about 10%, approximately 1,000 new AI-related positions are expected to be created.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
AI system task:
Why's our monitor labelling this an incident or hazard?
The announcement describes a concrete plan to use AI to perform tasks currently done by humans, directly leading to mass layoffs (harm to workers’ livelihoods). This is the use of an AI system causing realized (or imminently realized) harm to individuals, fitting the definition of an AI Incident under labor-rights and economic-harm criteria.[AI generated]
British Musicians Protest AI Training Copyright Revisions
Over 1,000 British musicians, including Kate Bush and Cat Stevens, released a silent album in protest of proposed UK copyright law changes. The revisions would allow tech companies to use artists' works for training AI models without compensation, raising significant ethical and legal concerns regarding intellectual property rights.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
AI system task:
Why's our monitor labelling this an incident or hazard?
The event involves AI systems in the context of training on copyrighted creative works, which is central to the dispute. However, no actual harm has yet occurred; the event is about the potential future harm to artists' rights and livelihoods if the law changes as proposed. This constitutes a plausible risk of harm due to AI system development and use, but no direct or indirect harm has materialized yet. Therefore, this is best classified as an AI Hazard, reflecting the credible risk posed by AI training practices under the proposed legal changes.[AI generated]
Parents Denounce AI-Generated Nude Images of Minors in Villablino
Parents of four minor female students at Instituto Obispo Argüelles in Villablino, León, have filed a complaint with Guardia Civil over the spread of nude images generated by AI. Authorities are investigating to determine who is behind the use and distribution of these non-consensual images.[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 used to generate manipulated images of minors, which were then circulated among students. This use of AI has directly led to harm in terms of privacy violations, potential psychological harm to the minors involved, and breaches of legal protections for children. The incident involves the use and misuse of AI-generated content causing real harm, fitting the definition of an AI Incident.[AI generated]
Delhi High Court Denies Early Hearing on Ban of DeepSeek AI Chatbot
The Delhi High Court refused an expedited hearing on a petition seeking to block DeepSeek, a Chinese-developed AI chatbot accused of leaks of sensitive personal and government data. The bench ruled no urgency, advising users to avoid the platform until the regular hearing, scheduled for April.[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 involves an AI system (DeepSeek chatbot) and a legal plea concerning its access, but there is no indication that the AI system has caused any direct or indirect harm yet. The court's refusal for early hearing suggests no urgent harm or incident has materialized. Therefore, this is a potential future risk scenario related to the AI system, but no harm has occurred or is imminent as per the article. This fits the definition of an AI Hazard, as the development or use of the AI system could plausibly lead to harm, but no incident has occurred yet.[AI generated]
Future Global Job Displacement Amid AI and Automation Advances
Multiple experts and studies warn that AI and automation could eventually displace 45–47% of global jobs. They highlight the risk of widespread unemployment and mental health issues, urging the implementation of digital literacy and retraining programs to mitigate the anticipated socioeconomic impacts.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Why's our monitor labelling this an incident or hazard?
The article clearly involves AI systems and their use in automation and generative AI, and it discusses the potential for widespread job losses and social harms due to these technologies. However, it does not report an actual incident of harm caused by AI, nor does it describe a specific hazard event where harm was narrowly avoided. The focus is on the plausible future risk of unemployment and social disruption due to AI and automation, making it a discussion of potential hazards at a broad societal level rather than a concrete AI Hazard event. It also includes recommendations and observations about responses to these challenges, but these do not constitute complementary information about a specific incident. Therefore, the article is best classified as an AI Hazard because it highlights credible potential harms from AI's impact on employment and society.[AI generated]
Romanian presidential candidate accuses rival of using ChatGPT to incite violence
Independent candidate Călin Georgescu has filed a criminal complaint against strategist Andrei Caramitru, alleging he used ChatGPT to produce content inciting violence and death threats. Georgescu cites penal code provisions on public incitement to violence, hate, and discrimination after a Facebook post generated thousands of reactions.[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 AI services to produce false evidence aimed at inciting violence, which is a direct harm to individuals and communities. The AI system's use in this context is linked to a violation of rights and potential physical harm, meeting the criteria for an AI Incident.[AI generated]
iPhone Voice Recognition Error Transcribes 'Racist' as 'Trump'
Apple's iPhone AI mis-transcribed 'racist' as 'Trump' during voice dictation, sparking controversy and viral reactions on TikTok. The company acknowledged the error, attributing it to phonetic overlap, and is working on a fix to prevent miscommunication and reputational harm.[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 voice-to-text dictation function involves an AI system (speech recognition model). The malfunction led to incorrect transcription, which caused social controversy and concern among users. Although no physical harm or legal violation is explicitly reported, the incident involves an AI system malfunction that has led to reputational and social harm (controversy and misinformation). This fits the definition of an AI Incident as the AI system's malfunction directly led to a harm to communities (social controversy and misinformation).[AI generated]
Romania to Deploy Combat-Proven AI V-BAT Drones in Black Sea
Romania will receive four MQ-35A V-BAT autonomous drones donated by the United States, capable of persistent AI-driven surveillance and targeting of Russian positions in the Black Sea. Already used in Ukraine to locate and destroy enemy air defenses, these drones strengthen Romanian naval reconnaissance with plans to acquire two additional systems.[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 MQ-35A V-BAT drones are AI systems as they perform autonomous navigation, data collection, and real-time decision-making in complex environments. Their use in targeting and destroying enemy air defense systems directly leads to harm to property and military assets, fulfilling the criteria for an AI Incident. The article reports actual use and realized harm, not just potential risk, so this is not an AI Hazard. It is not merely complementary information or unrelated news, as the AI system's use has directly led to harm in a military conflict context.[AI generated]
US Government Deepfake Incident Under Investigation
The US Department of Housing is investigating a hacking incident where internal screens briefly showed an AI-generated deepfake video of Donald Trump licking Elon Musk's bare feet with a 'Long Live The Real King' caption. Authorities are examining the misuse of taxpayer resources and the reputational impact on involved parties.[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 event involves an AI system generating a deepfake video, which is a clear example of AI-generated manipulated content. The video was displayed on internal government screens, which is a misuse of resources and could cause reputational harm to individuals and undermine trust in government operations. The harm is realized, not just potential, as the video circulated on social media and appeared on official screens, prompting an official investigation. Therefore, this meets the criteria for an AI Incident due to direct harm caused by the AI system's use.[AI generated]
Microsoft AI Incidents: Entra ID Disruption and OpenAI Exploit
Microsoft resolved a DNS authentication issue in its Entra ID service that disrupted Azure access due to a recent DNS change. Meanwhile, hackers identified as Storm-2139 bypassed security in Azure OpenAI tools, enabling the creation of harmful AI-generated content, including non-consensual celebrity images, highlighting vulnerabilities in its AI infrastructure.[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 event involves the malicious use of AI systems (Azure OpenAI) to produce harmful, non-consensual sexual images, which directly leads to harm to individuals and violations of rights. The hackers' actions represent misuse of AI capabilities causing real harm, qualifying this as an AI Incident. The involvement of AI systems in generating harmful content and the resulting harm to persons meet the criteria for an AI Incident rather than a hazard or complementary information.[AI generated]
AI Flags Actress for Financial Discrepancies
The Greek tax authority's AI system detected a significant mismatch between the declared income and bank transactions of a well-known actress. This discrepancy has spurred a potential legal investigation into tax evasion or money laundering, raising concerns about the protection of fundamental rights.[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 event involves the use of an AI system by a government tax authority to detect possible tax fraud by analyzing financial data. The AI system's use directly led to the identification of a significant discrepancy, triggering an official investigation, which constitutes a harm related to legal compliance and potential violation of financial laws. Therefore, this qualifies as an AI Incident because the AI system's use has directly led to a legal and regulatory harm scenario.[AI generated]
Controversial AI Surveillance Demo Sparks Backlash
A YC-backed startup, Optifye.AI, demoed an AI-driven system that monitors factory workers with machine vision and harsh performance critiques, labeling one worker as 'Number 17'. The dehumanizing demonstration raised human rights concerns and led to widespread criticism, prompting Y Combinator to remove the video from its platforms.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Business function:
AI system task:
Why's our monitor labelling this an incident or hazard?
The AI system (Optifye.ai) uses machine vision to monitor workers' hand movements and output, providing supervisors with detailed productivity data. This AI-enabled surveillance directly affects workers by enabling micromanagement and punitive actions based on AI assessments, which can harm workers' dignity, privacy, and labor rights. The article describes actual use and deployment intentions, indicating realized harm rather than just potential risk. Therefore, this qualifies as an AI Incident due to violations of labor rights and harm to workers caused by the AI system's use.[AI generated]
Microsoft Copilot Exposes Private GitHub Data
An Israeli cybersecurity firm discovered that Microsoft's Copilot AI tool can access historical data from private GitHub repositories, exposing sensitive information and intellectual property. Over 16,000 organizations, including Microsoft, AWS, and Google, are affected, raising concerns over data security and privacy.[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 event involves an AI system (Microsoft Copilot) that accesses and exposes sensitive data due to its reliance on cached public GitHub repositories. This has directly led to a data breach risk affecting thousands of organizations, including major companies. The harm includes potential violations of intellectual property rights and privacy, which fits the definition of an AI Incident as the AI system's use has directly led to harm or risk of harm. The issue is not merely potential but ongoing, as Copilot can still access sensitive data despite mitigation efforts, thus constituting an AI Incident rather than a hazard or complementary information.[AI generated]
Volatus and Ondas Partner on AI-Powered Autonomous Drones for Border Surveillance
Volatus Aerospace and Ondas Holdings’ subsidiary American Robotics have teamed up to market the Optimus System, a fully autonomous AI-driven drone platform providing persistent 24/7 border surveillance, intelligence gathering, and emergency response. While enhancing national security, the deployment of these drones carries credible risks of privacy violations, misuse, and rights infringements.[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 event involves the use of an AI system (the Optimus System autonomous drone platform) for surveillance and security purposes. While the system is designed to enhance national security and border monitoring, the article does not report any actual harm or incidents resulting from its deployment. However, the deployment of autonomous drones for border surveillance carries plausible risks of harm, such as privacy violations, misuse, or unintended consequences affecting human rights or communities. Since no harm has yet occurred or been reported, but plausible future harm could arise from the use of such AI-enabled autonomous surveillance systems, this event qualifies as an AI Hazard rather than an Incident or Complementary Information.[AI generated]
Global Surge in AI-Driven Deepfake Scams
A global study by Hiya highlights a rising threat of AI-driven deepfake phone scams across Canada, the US, and the UK. Fraudsters use deepfake technology to impersonate trusted voices—most notably through fake Amazon calls in Canada and Medicare or tax scams elsewhere—leading to significant financial losses for consumers.[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 AI-generated voice impersonation scams (deepfake fraud calls) that have resulted in actual financial harm to consumers, with reported monetary losses. The AI system's use in creating convincing fraudulent calls is central to the harm caused, fulfilling the criteria for an AI Incident due to direct harm to persons (financial injury).[AI generated]
Morgan & Morgan Lawyers Sanctioned for AI-Generated Fake Case Citations
In Wyoming, Judge Kelly H. Rankin sanctioned Morgan & Morgan lawyers for submitting AI-generated fake case citations in a lawsuit against Walmart. Rudwin Ayala lost his pro hac vice admission and was fined $3,000; two colleagues each paid $1,000.[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 lawyers used an AI system to generate legal case citations that were fabricated ('hallucinated'), which were then submitted in official court documents. This misuse of AI directly led to legal sanctions and fines, demonstrating a clear harm caused by the AI system's outputs. The event involves the use and misuse of AI, resulting in a breach of legal and ethical standards, which fits the definition of an AI Incident due to violation of applicable law and harm to the legal process.[AI generated]
AI Misuse Creates Fake Explicit Content Targeting Porsche Sales Champion
Qingdao Porsche sales champion Mou Qianwen reported that AI technology was used to maliciously manipulate her photos into fake explicit images and videos posted on adult websites. She confirmed the content was unauthorized, filed a police report, and condemned the misuse of AI for defamation.[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 describes how AI technology was used to generate fake pornographic images of a person by maliciously manipulating her photos. This misuse of AI has directly led to harm in the form of reputational damage and personal rights violations. Therefore, it qualifies as an AI Incident due to the realized harm caused by the AI system's malicious use.[AI generated]
Faulty AI Facial Recognition Leads to Wrongful Arrest in Detroit
Detroit police wrongly arrested LaDonna Crutchfield after a facial recognition error misidentified her as a shooting suspect. Despite police claims contesting AI involvement in one instance, a lawsuit argues that reliance on such technology violated her rights and caused significant emotional distress.[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 event describes a wrongful arrest based on a false facial recognition match, which is an AI system used in law enforcement. The harm includes emotional distress and violation of rights due to misidentification. The AI system's malfunction or misuse directly led to this harm, fulfilling the criteria for an AI Incident. Despite police denial, the lawsuit and pattern of similar cases support the involvement of AI facial recognition causing harm.[AI generated]