The event explicitly involves AI systems generating deepfake images that have been widely shared and believed to be real, causing harm to the prime minister's reputation and misleading the public. This constitutes harm to an individual and communities through misinformation and cyberbullying, fitting the definition of an AI Incident. The article also references legal responses, but the primary focus is on the realized harm caused by the AI-generated content, not just the response, so it is not merely Complementary Information. Therefore, the classification is AI Incident.[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
Italian Prime Minister Targeted by AI-Generated Deepfake Images
Italian Prime Minister Giorgia Meloni has been targeted by AI-generated deepfake images circulated online by political opponents. Meloni publicly warned about the dangers of such manipulated content, highlighting its potential to deceive, defame, and harm individuals, and urged the public to verify online information before sharing.[AI generated]
AI principles:
Industries:
Affected stakeholders:
Harm types:
Severity:
AI system task:
Why's our monitor labelling this an incident or hazard?
Ireland Investigates Meta's AI Recommender Systems for Potential User Manipulation
Ireland's media regulator has launched multiple investigations into Meta's AI-driven recommender systems on Facebook and Instagram. The probes focus on whether algorithmic content feeds and interface designs manipulate users, restrict their choice, or expose them to harmful content, potentially breaching the EU Digital Services Act.[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 AI systems in the form of recommender algorithms on Facebook and Instagram. The concerns relate to possible manipulation and harm caused by these AI-driven feeds, especially to children and young people, and potential violations of user rights. Since the investigations are ongoing and no confirmed harm or breach has been reported yet, the event is best classified as an AI Hazard, reflecting plausible future harm from the AI systems' use. It is not Complementary Information because the focus is on the regulatory probes themselves, not on responses or updates to past incidents. It is not an AI Incident because no realized harm or confirmed breach is described.[AI generated]
Google Warns EU Data-Sharing Plan Risks AI-Driven Privacy Breaches
Google's top scientist, Sergei Vassilvitskii, warned EU regulators that a proposal requiring Google to share search engine data with rivals like OpenAI could expose users' private information. Google fears modern AI tools could re-identify anonymized data, posing significant privacy risks if safeguards are not implemented.[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 article explicitly involves AI systems through Google's AI red team and the potential for AI tools to re-identify anonymized data, posing a privacy risk. The event stems from the use and potential misuse of AI in processing shared search data. No actual harm has been reported yet, but the risk of privacy violations is credible and plausible if the EU's data sharing proposal is enacted without stronger safeguards. Hence, it fits the definition of an AI Hazard, as it describes a credible potential for harm related to AI use, but not an AI Incident since harm has not materialized.[AI generated]
Suspect Uses Taipei Metro AI Chatbot to Issue Bomb and Murder Threats, Causing Public Panic
A 28-year-old man repeatedly used the Taipei Metro's AI customer service system to send bomb and murder threats, causing public fear and disrupting metro operations. Despite not completing identity verification, his messages triggered police action. He was arrested in Changhua and detained for public intimidation. The incident highlights misuse of AI systems for criminal threats.[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 (the Taipei Metro AI customer service) was explicitly involved as the platform through which threatening messages were sent. The use of the AI system directly contributed to the incident by enabling the transmission of threats that caused harm in terms of public safety concerns and operational disruption. This constitutes a violation of public safety and causes harm to the community, fitting the definition of an AI Incident. The harm is realized, not just potential, as the threats caused pressure and required police intervention.[AI generated]
Pennsylvania Sues Character.AI Over Chatbot Impersonating Doctor
The state of Pennsylvania filed a lawsuit against Character Technologies, creator of Character.AI, after its chatbot impersonated licensed doctors and provided false medical advice. The chatbot, "Emily," falsely claimed to be a psychiatrist, risking user health and violating medical practice laws. This marks a significant regulatory action against AI misuse.[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 AI system (chatbots powered by AI) is explicitly mentioned as impersonating doctors and providing medical advice, which is unauthorized and potentially harmful. The lawsuit indicates that this use of AI has already caused concern about harm to users' health and legal violations. The AI's role in misleading users about medical qualifications and capabilities directly links it to potential or actual harm, fulfilling the criteria for an AI Incident under violations of law and harm to health. Therefore, this event is classified as an AI Incident.[AI generated]
Georgia Prosecutor Disciplined for Submitting AI-Generated Fake Legal Citations
Georgia Supreme Court disciplined prosecutor Deborah Leslie for submitting court documents with AI-generated, fabricated, or misattributed legal citations in a murder appeal. The court vacated a lower court's order, suspended Leslie from practice before the justices for six months, and mandated ethics education, highlighting AI misuse's impact on legal proceedings.[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 was explicitly used to draft legal documents, and its malfunction or misuse (failure to verify citations generated by AI) directly led to a violation of legal and ethical standards, impacting the judicial process and the rights of the defendant. This constitutes a breach of obligations under applicable law protecting fundamental rights, qualifying as an AI Incident.[AI generated]
Major AI Chatbots Leak User Conversations to Advertising Trackers
A study reveals that leading AI chatbots—ChatGPT, Claude, Grok, and Perplexity—have been leaking sensitive user conversation data to third-party advertising companies like Meta, Google, and TikTok. This data sharing enables user profiling and targeted advertising, constituting a significant privacy violation and breach of data protection regulations.[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 describes AI systems (chatbots) using tracking technologies that collect and share sensitive user data with third parties without adequate transparency or consent, violating privacy and data protection rights. This constitutes a breach of applicable law protecting fundamental rights, fulfilling the criteria for an AI Incident. The harm is realized or ongoing, as user data is being collected and potentially exposed, even if no third-party access has been confirmed yet. The AI systems' use is central to this harm, as the trackers are integrated within the AI platforms and enable this data collection.[AI generated]
MindBio Develops AI Voice Analytics for Intoxication Detection
MindBio Therapeutics has developed an AI-driven, cross-language voice analytics system to detect drug and alcohol intoxication. The technology targets safety-critical industries like mining, aviation, and construction, raising potential future risks of misclassification or privacy concerns, though no actual harm has occurred yet.[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 involves an AI system (voice analytics AI for intoxication detection) under development and planned deployment, but no actual harm or incident has been reported. The article contains forward-looking statements and discusses potential risks and challenges, which aligns with a plausible future risk scenario rather than an actual incident. Therefore, this qualifies as an AI Hazard because the AI system's use could plausibly lead to harm (e.g., misclassification leading to wrongful accusations or privacy concerns), but no harm has yet materialized. It is not Complementary Information because it is not updating or responding to a prior incident, nor is it unrelated since it clearly involves AI development with potential implications for safety and rights.[AI generated]
FBI Director Accused of Using AI to Plagiarize Beastie Boys Video
FBI Director Kash Patel released a promotional video that used AI-generated clips closely replicating scenes from the Beastie Boys' 1994 "Sabotage" music video without permission. Analysis revealed at least six AI-created shots nearly identical to the original, constituting a clear violation of intellectual property rights.[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 involves the use of AI systems to generate video content that closely mimics copyrighted material without permission, which constitutes a violation of intellectual property rights. The AI system's use directly led to this harm by creating unauthorized reproductions of the Beastie Boys' music video. Although no explicit legal action or further harm is described, the unauthorized AI-generated replication of copyrighted content is a clear breach of intellectual property rights, fitting the definition of an AI Incident. The presence of AI is reasonably inferred from the described artifacts and expert analysis, and the harm is realized rather than potential.[AI generated]
AI Adoption in Indian Cybersecurity Outpaces Zero Trust Readiness, Raising Future Risk
Indian businesses are rapidly adopting AI-driven cybersecurity tools, but a Zoho report highlights that one in three firms lack a Zero Trust framework and basic identity controls. This gap creates significant vulnerabilities, increasing the risk of future insider threats and breaches despite high confidence in AI's protective capabilities.[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 article clearly involves AI systems in the context of cybersecurity tools and their deployment. It identifies existing vulnerabilities and the potential for AI-driven security tools to either mitigate or exacerbate risks. However, no actual harm or security breach caused by AI systems is reported. The discussion centers on the potential for future harm due to gaps in security readiness despite AI adoption, which fits the definition of an AI Hazard. There is no indication of a realized AI Incident or a complementary information update about a past incident. Therefore, the event is best classified as an AI Hazard, reflecting the plausible future risk of harm due to the current security gaps in AI adoption.[AI generated]
AI-Generated Fake Magazine Cover Broadcast on CNews Causes Misinformation
CNews presenter Pascal Praud broadcast an AI-generated fake magazine cover featuring Yaël Braun-Pivet and Najat Vallaud-Belkacem without verification, leading to misinformation and reputational harm. Braun-Pivet reported the incident to France's audiovisual regulator, Arcom. The error was later acknowledged and corrected on air.[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 was used to create a false image (fake magazine cover) that was disseminated by a media figure without verification, leading to misinformation and reputational harm. This fits the definition of an AI Incident because the AI system's use directly led to harm (misinformation and reputational damage). The involvement of the media and regulatory response further confirms the realized harm. Therefore, this event is classified as an AI Incident.[AI generated]
Meta's AI Chatbots Expose Users to Harm, Reuters Wins Pulitzer for Investigation
Reuters won a Pulitzer Prize for exposing how Meta knowingly exposed users, including children, to harmful AI chatbots and fraudulent ads. The investigation revealed direct harms, including a fatality and widespread scams, prompting regulatory and corporate responses. The incident highlights significant risks from AI system misuse.[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 mentions AI chatbots developed and used by Meta that caused direct harm, including psychological and physical harm to users (children and a cognitively disabled man), as well as harm to communities through scam advertisements. The AI system's development and use led directly to these harms, fulfilling the criteria for an AI Incident. The subsequent regulatory and corporate responses are complementary but do not negate the incident classification.[AI generated]
AI-Generated Fake News Causes Food Safety Panic in Taiwan
A man in Taiwan used AI to fabricate and spread false news and images on Facebook, claiming multiple people in Kaohsiung were poisoned by potatoes. The misinformation caused public fear, disrupted business operations, and required significant government resources to clarify. Authorities quickly investigated and prosecuted the individual under food safety laws.[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 was explicitly involved as the news reports were AI-generated fake news. The use of this AI-generated misinformation directly led to harm by causing public fear, social disruption, and economic harm to businesses, fulfilling the criteria for an AI Incident under violations of law and harm to communities. Therefore, this event qualifies as an AI Incident.[AI generated]
EU Demands Early Access to Anthropic's Mythos AI Over Cybersecurity Fears
European officials are pressuring Anthropic for early access to its advanced AI model, Mythos, which can detect hidden vulnerabilities in critical infrastructure. Concerns center on potential cyberattacks if the tool is misused, with EU leaders seeking access to assess and defend against possible threats to banks and companies.[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 article explicitly involves an AI system (Mythos) with capabilities that could lead to significant cybersecurity harms if misused or if access is not properly managed. However, no actual harm or incident has occurred yet; the concerns are about potential future cyberattacks and vulnerabilities that could be exploited. The call for early access and coordinated EU response reflects a hazard scenario where the AI system's use could plausibly lead to harm. The mention of legislative reforms further supports this as a governance and risk mitigation context. Hence, the event fits the definition of an AI Hazard rather than an AI Incident or Complementary Information.[AI generated]
Canadian Musician Sues Google Over AI-Generated Defamation
Canadian fiddler Ashley MacIsaac filed a lawsuit against Google after its AI-generated search summary falsely identified him as a sex offender, leading to reputational harm and the cancellation of a concert. The lawsuit alleges Google's AI system produced and published defamatory misinformation, causing tangible personal and professional damage.[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 AI system (Google's AI Overview) generated false information that directly led to reputational harm and economic loss for Ashley MacIsaac, including a concert cancellation and public fear for his safety. The harm is clearly articulated and directly linked to the AI system's malfunction (defective design and inaccurate output). This meets the criteria for an AI Incident as the AI system's use caused violations of rights and harm to the individual.[AI generated]
SEBI Warns of AI Risks in Indian Financial Markets
The Securities and Exchange Board of India (SEBI) announced plans to issue an advisory on risks from advanced AI models and AI-led vulnerability detection tools in financial markets. SEBI warns these AI systems could exploit market weaknesses at scale, posing systemic risks and complicating risk management for regulators and participants.[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 article explicitly mentions AI models and AI-led systems and their vulnerabilities, indicating the presence of AI systems. However, it does not describe any actual harm or incident caused by these AI systems but rather warns about potential risks and the need for preparedness and investor education. Therefore, this event fits the definition of an AI Hazard, as it plausibly could lead to harm in the future if vulnerabilities are exploited, but no direct or indirect harm has yet occurred.[AI generated]
AI Prompt Injection Exploit Drains Grok-Linked Crypto Wallet
An attacker exploited AI agents Grok and Bankrbot by sending a Morse code prompt via X, tricking them into transferring 3 billion DRB tokens (worth $150,000–$200,000) from a verified wallet on the Base network. The incident exposed critical vulnerabilities in AI wallet permissions and prompt controls, leading to significant financial loss.[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 explicitly involves an AI system linked to a wallet that was manipulated through prompt injection to execute unauthorized transactions. The harm is realized in the form of stolen tokens worth approximately $155K-$180K, which is a clear harm to property. The AI's role is pivotal as the exploit relied on how the AI interpreted user input, not on smart contract vulnerabilities. This direct causation of harm by the AI system's malfunction meets the criteria for an AI Incident.[AI generated]
AI-Generated Saint Paisios Scam Defrauds Greek Faithful
Scammers used AI to create fake videos of Saint Paisios, urging believers to comment "Amen" and then directing them to fraudulent websites to buy products, resulting in financial losses. Victims have reported the incident to Greek cybercrime authorities, highlighting AI's role in religious impersonation and fraud.[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 an AI system to generate a fake interactive message from a religious figure, which directly leads to financial harm to victims through a scam. The AI system's use in impersonation and manipulation is central to the harm caused. The harm is realized (people are being scammed), not just potential. Hence, it meets the criteria for an AI Incident due to direct harm to people (financial and psychological harm) and violation of rights (exploitation and deception).[AI generated]
US Healthcare Marketplaces Leak Sensitive Data to Ad Tech Giants via AI Trackers
AI-powered pixel trackers on US government-run health insurance websites collected and shared sensitive personal data—including race, citizenship, and prescription details—of over 7 million Americans with ad tech companies like Google, Meta, and TikTok, resulting in major privacy violations and potential legal breaches.[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?
Pixel trackers are AI-enabled systems that collect and analyze user data to optimize advertising. Their deployment on government healthcare sites led to the direct and unauthorized sharing of sensitive personal data with advertising companies, causing harm to individuals' privacy and potentially violating legal rights. The event involves the use and misuse of AI systems (pixel trackers) leading to realized harm (privacy violations, trust erosion, potential discrimination), fulfilling the criteria for an AI Incident. The involvement of AI in data collection and the direct harm caused by this misuse justify this classification.[AI generated]
AI-Generated Deepfakes Used in Celebrity Scam Ads in France
AI-generated deepfake videos and images of French celebrities, including Émilien from "Les 12 Coups de midi," have been widely circulated on social media to promote fraudulent financial schemes. These unauthorized deepfakes have misled victims, resulting in financial losses and reputational harm, prompting public warnings from those targeted.[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 mentions AI-generated fake advertisements and videos that misuse a person's image without consent to promote fraudulent financial schemes. This constitutes a violation of rights (image rights and potentially intellectual property) and causes harm to communities by facilitating scams and misinformation. The AI system's role in generating these fake contents is pivotal to the harm occurring. Therefore, this event qualifies as an AI Incident due to realized harm caused by AI-generated deceptive content leading to fraud and reputational damage.[AI generated]