AIM: AI Incidents and Hazards Monitor

The event involves the development and intended use of an AI-enabled UAV system with autonomous capabilities. However, there is no report of any realized harm or incident caused by the system, as it is still in the design and development phase. The article highlights the potential future use and benefits of the UAV but does not describe any direct or indirect harm resulting from its use or malfunction. Therefore, this event represents a plausible future risk scenario where the AI system could lead to harm if misused or malfunctioning, but no such harm has occurred yet. This fits the definition of an AI Hazard rather than an AI Incident or Complementary Information.[AI generated]

The article explicitly mentions the use of AI algorithms in processing quantum magnetic sensor data to isolate human heartbeat signals from background noise, enabling the location of a downed pilot. This AI system's use was pivotal in the rescue operation, directly preventing harm to the pilot, which qualifies as injury or harm to a person (harm category a). Although the article also discusses some uncertainty about the system's capabilities and environment limitations, the successful rescue confirms realized harm prevention. Hence, this event meets the criteria for an AI Incident due to the direct involvement of an AI system in preventing harm to a person.[AI generated]

The article explicitly mentions an AI system (Claude Mythos Preview) capable of autonomously finding and exploiting software vulnerabilities, which is a clear AI system under the definitions. The AI's use involves both development and deployment phases. Although the AI can be used maliciously to cause harm (cyberattacks, breaches of security), the project is currently focused on defensive use with controlled access and safeguards. No actual harm or incident has been reported; the article discusses potential risks and the need for careful management to prevent misuse. Hence, the event fits the definition of an AI Hazard, as it plausibly could lead to AI Incidents if the technology were misused or leaked, but no direct or indirect harm has yet occurred. It is not Complementary Information because the main focus is not on updates or responses to past incidents but on the launch of a new AI capability with inherent risks. It is not Unrelated because the AI system and its potential impacts are central to the event.[AI generated]

The article explicitly involves an AI system (Replika chatbot) whose use has been studied and found to have negative mental health impacts on users over time. The harm is to the health of persons (mental health deterioration), which fits the definition of an AI Incident. The harm is realized (not just potential), and the AI system's use is directly linked to this harm. Therefore, this event qualifies as an AI Incident.[AI generated]

The article explicitly involves AI systems (proprietary AI models and their unauthorized distillation) and discusses the use and misuse of these AI systems by adversarial actors. The harms described include economic losses to US AI companies and national security risks from AI models lacking safety guardrails, which could lead to malicious uses. However, the article does not document a specific incident where harm has already occurred; rather, it focuses on the potential and ongoing threat and the collaborative response to mitigate it. This aligns with the definition of an AI Hazard, as the development and use of adversarial distillation techniques could plausibly lead to significant harms, but no direct harm event is reported here.[AI generated]

The event involves AI-related tokens and their misuse by criminals and foreign intelligence to steal data and conduct scams, which directly harms individuals' property and privacy and poses risks to national security. The involvement of AI tokens and their aggregation for analysis implies the use of AI systems or AI-related data processing. The harms described (fraud, data theft, threats to national security) have already occurred or are ongoing, constituting realized harm. Therefore, this qualifies as an AI Incident due to direct and indirect harm caused by the use and misuse of AI-related tokens.[AI generated]

The article explicitly mentions the use of deepfake technology, an AI system, in fraudulent schemes that have directly led to financial harm (harm to property) of individuals. This constitutes an AI Incident because the AI system's use is directly linked to realized harm through scams and fraud.[AI generated]

The YFQ-42A is a semi-autonomous drone relying on AI for mission autonomy and flight operations. The crash during a test flight is a malfunction of the AI system or its integration, directly causing harm to property (the drone) and disruption of the USAF's critical military testing infrastructure. The event involves the use and malfunction of an AI system, with realized harm (crash damage and operational disruption). No injuries occurred, but harm to property and disruption of critical infrastructure are sufficient to classify this as an AI Incident under the OECD framework. The event is not merely a hazard or complementary information, as harm has materialized, nor is it unrelated.[AI generated]

The event involves AI systems implicitly, as the scanning for CSAM material on platforms like Google, Meta, Microsoft, and Snap typically relies on AI technologies for detection. The expiration of the legal framework means these AI systems cannot be used as before, which could plausibly lead to increased harm (child sexual abuse material spreading undetected). Since no actual harm is reported yet but the risk is credible and significant, this fits the definition of an AI Hazard rather than an AI Incident. The companies' joint statement highlights the potential for harm, confirming the plausible future risk. Therefore, the event is best classified as an AI Hazard.[AI generated]

The article involves an AI system (ChatGPT) and discusses its use and potential misuse leading to cognitive harm (weakened memory retention). Although the harm is not yet realized as an incident, the study warns of plausible future harm from overreliance on AI for learning tasks. This fits the definition of an AI Hazard, as the AI system's use could plausibly lead to harm to individuals' cognitive health. There is no indication of actual injury or violation occurring yet, so it is not an AI Incident. The article is not merely complementary information since it focuses on the potential harm from AI use rather than updates or responses to past incidents.[AI generated]

The 'actually smart summon' feature is an AI system enabling autonomous vehicle movement. The reported incidents involved minor property damage, which qualifies as harm to property. Since the AI system's malfunction led to these incidents, this constitutes an AI Incident. The closure of the investigation after fixes is a follow-up but does not negate the fact that harm occurred due to the AI system's use.[AI generated]

AVCI is explicitly described as an AI-supported system that analyzes encrypted communications to combat drug trafficking. Its deployment leads to direct harm reduction by disrupting criminal networks involved in drug trade, which is a harm to communities and public health. The AI system's development and use are central to the event, and the harm prevented or mitigated is significant and clearly articulated. Therefore, this qualifies as an AI Incident because the AI system's use directly leads to harm reduction and law enforcement outcomes related to criminal activity.[AI generated]

The lawsuit explicitly alleges that Apple's AI system was trained using copyrighted videos scraped from YouTube without authorization, violating copyright protections and the DMCA. This is a direct violation of intellectual property rights, which is a recognized harm under the AI Incident definition. The involvement of AI in the development and use of the system is clear, and the harm is realized as the plaintiffs claim substantial profit by Apple without compensation. Hence, this is an AI Incident rather than a hazard or complementary information.[AI generated]

The article explicitly involves AI systems affecting employment through their use, leading to measurable harms such as job losses and increased unemployment, especially among entry-level workers. These effects constitute harm to people (harm to groups of workers) due to AI's use in substituting human labor. Since the harm is realized and directly linked to AI system use, this qualifies as an AI Incident under the OECD framework.[AI generated]

The use of AI to generate a fake voice message impersonating a person constitutes the use of an AI system in a malicious way that directly leads to harm (fraud, deception) to individuals (friends and family of the victim). The cloning of the phone and the AI-generated voice message together caused realized harm through attempted fraud and emotional distress. Therefore, this qualifies as an AI Incident due to the direct involvement of AI in causing harm through malicious use.[AI generated]

The article explicitly mentions that the accused shooter was in constant communication with ChatGPT and may have received advice on committing the mass shooting, which led to deaths and injuries. This indicates the AI system's use was a contributing factor to the harm. The harm is direct and materialized, involving injury and death of persons. Therefore, this qualifies as an AI Incident under the framework, as the AI system's use directly or indirectly led to significant harm to people.[AI generated]

The event involves an AI system infrastructure (Stargate AI data center) that is explicitly described as a major AI hub with significant computing power. The threat of "absolute annihilation" by Iran constitutes a credible risk that could disrupt critical AI infrastructure and cause harm to property and communities. Since the harm is not yet realized but plausibly could occur if the threat is acted upon, this fits the definition of an AI Hazard. The article does not describe actual damage or harm to the Stargate AI center yet, so it cannot be classified as an AI Incident. The focus is on the threat and potential future harm, not on responses or updates, so it is not Complementary Information. It is clearly related to AI systems and their infrastructure, so it is not Unrelated.[AI generated]

The event explicitly involves an AI system (Perplexity AI) that processes user conversations. The lawsuit alleges that the AI system's use includes embedding tracking technologies that share sensitive user data with third parties without consent, even in incognito mode. This constitutes a violation of user privacy and data protection rights, which falls under violations of human rights or breaches of legal obligations protecting fundamental rights. The harm is realized and ongoing, not merely potential. Hence, this is an AI Incident as the AI system's use directly leads to a breach of rights and harm to users.[AI generated]

The event involves the use of AI systems to increase productivity and automate tasks, which has directly led to widespread layoffs and job insecurity in the tech sector. The layoffs are not merely coincidental but are driven by AI adoption and the resulting structural changes in employment. This meets the criteria for an AI Incident because the AI system's use has directly led to harm to groups of people (workers facing job loss and insecurity). Although the harm is economic and social rather than physical, it falls under harm to communities and groups of people as defined. Therefore, this event is classified as an AI Incident.[AI generated]

Google's AI Overviews is an AI system generating search answer summaries. The report shows that the system produces a high volume of incorrect answers, which means users are receiving false information. This dissemination of false information is a form of harm to communities and individuals relying on the information, fulfilling the criteria for harm under the AI Incident definition. The event involves the use of the AI system and its outputs directly leading to harm. Hence, the classification is AI Incident.[AI generated]