AI-Assisted Code Changes Cause Major Outages at Amazon

The outages are directly linked to the use and deployment of AI systems (Gen-AI assisted changes and AI coding assistants) that led to service disruptions affecting customers' ability to use Amazon's platform. This constitutes harm to users by disrupting access to critical online retail services, which falls under harm to communities and property (economic harm). Therefore, this qualifies as an AI Incident due to the realized harm caused by AI system use and malfunction.

The article explicitly mentions that some outages were tied to AI-assisted coding errors, indicating AI system involvement in the development or deployment phase. The outages directly led to harm by disrupting Amazon's website and app functionality, affecting users' ability to shop and access account information, which constitutes harm to critical infrastructure and users. The company's response meeting is a reaction to these incidents, but the primary event is the AI-related outages causing real harm. Hence, this is classified as an AI Incident.

The article explicitly states that Amazon's AI coding assistant Q was one of the primary contributors to a major incident on March 2 that caused nearly 120,000 lost orders and 1.6 million website errors. This is a direct link between the AI system's use and significant harm to customers and the company's operations, fulfilling the criteria for an AI Incident. The harm is material and has already occurred, including lost orders and service outages. Although Amazon is implementing new safeguards, the article's main focus is on the realized harm and the AI system's role in causing it, not just potential future harm or complementary information about responses. Therefore, the event is best classified as an AI Incident.

The article explicitly mentions AI-assisted coding tools causing outages and disruptions at Amazon, including a major 13-hour outage at AWS. These outages represent disruption of critical infrastructure (cloud services), which is a defined harm under AI Incident criteria. The AI system's malfunction and use are directly linked to these harms. The organizational response to restrict AI-assisted code pushes is a mitigation measure but does not negate the fact that harm occurred. Hence, this is an AI Incident.

The report explicitly states that AI-generated code changes caused outages lasting several hours on Amazon's e-commerce platform and AWS services, disrupting critical infrastructure and customer access. This meets the definition of an AI Incident because the AI system's use directly led to harm (service disruption). The involvement of AI coding tools in generating faulty updates and deleting production environments is clear. The harm is realized, not just potential, so it is not an AI Hazard. The article also mentions internal responses, but the main focus is on the harm caused by AI use, not just complementary information.

The article explicitly mentions that AI-assisted changes, particularly those involving generative AI coding tools, contributed to software deployment errors that caused outages lasting nearly six hours. This disruption affected the operation of Amazon's ecommerce infrastructure, which qualifies as harm under the disruption of critical infrastructure category. Since the AI system's use directly or indirectly led to this harm, the event qualifies as an AI Incident rather than a hazard or complementary information.

The article explicitly mentions AI coding tools (generative AI) being used in software development, which caused significant outages in Amazon's services. These outages disrupted critical infrastructure (online retail and cloud services), directly harming users and the business. The AI system's malfunction and the company's acknowledgment of insufficient safeguards confirm the AI system's role in causing harm. Hence, this is an AI Incident rather than a hazard or complementary information.

The article involves AI systems in the context of AI-assisted coding tools potentially linked to outages, but no confirmed harm or direct causation by AI is established. The focus is on Amazon's denial of AI's role and discussion of organizational factors, making this a case of complementary information about AI's impact and governance rather than an incident or hazard. There is no new harm or credible risk of harm directly attributed to AI in this report.

The article explicitly mentions that outages and malfunctions in Amazon's services were traced back to code developed by generative AI coding assistants. These outages disrupted the availability of critical infrastructure services, which fits the definition of harm under category (b) "Disruption of the management and operation of critical infrastructure." The AI system's malfunction (buggy AI-generated code) directly caused these outages. The organizational response to require senior engineer approval is complementary information but does not negate the fact that harm occurred. Hence, this is an AI Incident.

The article explicitly states that generative AI coding tools were used to produce software changes that caused outages and service disruptions at Amazon and AWS. These outages harmed users by preventing access to services and completing purchases, which is a direct harm caused by the AI system's outputs. The involvement of AI in the development and deployment of faulty code changes that led to these incidents meets the criteria for an AI Incident. The company's response to require senior engineer oversight is a mitigation measure but does not negate the fact that harm has already occurred due to AI-assisted code deployment.

The article explicitly states that generative AI tools used in code generation and review have led to software deployment errors causing significant outages at Amazon, directly impacting users and service availability. This meets the definition of an AI Incident because the AI system's use has directly led to harm through service disruption, which affects critical infrastructure and harms communities relying on Amazon's services. The event is not merely a potential risk or a complementary update but a concrete incident with realized harm linked to AI system use.

The article explicitly mentions outages at Amazon Web Services linked to AI-assisted code changes, indicating that the AI system's outputs contributed to service disruptions. This is a clear example of harm to critical infrastructure management and operation caused by the use of an AI system. The response by Amazon to require senior engineer sign-off is a mitigation measure but does not negate the fact that harm occurred. Hence, the event meets the criteria for an AI Incident.

The article explicitly mentions that AI-assisted coding tools contributed to outages causing millions of lost orders and errors affecting customers globally. This is a direct harm linked to the AI system's use and malfunction. The harm includes disruption of critical infrastructure related to retail operations and harm to customers' experience and property (orders). The company's corrective measures confirm the AI system's role in causing the incident. Hence, the event meets the criteria for an AI Incident.

The article explicitly states that AI coding tools were primary contributors to outages that caused significant disruption and economic harm to users and the company. The AI systems' use and malfunction directly led to these harms. Therefore, this qualifies as an AI Incident because the AI system's use and malfunction have directly led to realized harm (service outages, lost orders, and user impact). The company's response and policy changes are complementary information but do not change the classification of the event as an AI Incident.

The article explicitly mentions an AI system (AI coding assistant Q) whose use in generating code directly contributed to technical failures causing harm to customers (incorrect delivery timelines, website errors, cancelled orders). This constitutes harm to communities and property (consumer trust, service reliability). The AI system's malfunction or misuse is a direct factor in these incidents. Therefore, this qualifies as an AI Incident rather than a hazard or complementary information.

The article explicitly mentions AI systems (AI coding tools, generative AI) causing system failures that disrupted Amazon's ecommerce and cloud services, which are critical infrastructure. The harm is realized and significant, as customers were unable to use key services for hours. This fits the definition of an AI Incident because the AI system's use directly led to harm (disruption of critical infrastructure). The company's response and mitigation efforts are described but do not change the classification, as the primary focus is on the incidents themselves and their impact.

The article explicitly mentions AI coding tools as contributing factors to outages that caused service disruptions, which qualify as harm to critical infrastructure. The AI system's use in code deployment led to incidents with a high blast radius affecting availability. Although not all outages were caused by AI, some were directly linked to AI-assisted code changes. Therefore, this qualifies as an AI Incident due to realized harm from AI system use in critical infrastructure management.

The outages caused by AI-assisted code changes directly led to harm in the form of disruption to critical infrastructure (Amazon's ecommerce platform and AWS services), fulfilling the criteria for an AI Incident. The AI system's involvement is in the use of AI-assisted code generation that contributed to erroneous deployments causing service outages. The harm is realized and significant, and the response involves human oversight to mitigate further incidents. Therefore, this event qualifies as an AI Incident.

The AI system (Kiro AI coding tool) was actively used to update critical infrastructure code, and its autonomous actions (deleting and recreating environments) directly caused large-scale outages and operational harm. This fits the definition of an AI Incident because the AI system's use directly led to significant harm (disruption of critical infrastructure and economic loss). The company's response to mitigate future risk is complementary information but does not negate the incident classification. The harm is realized and substantial, not merely potential.

The article explicitly mentions that an AI coding bot was responsible for causing service outages at Amazon, which are critical incidents affecting the availability of services. This is a direct harm to the operation of a critical infrastructure component (Amazon's e-commerce platform). The AI system's malfunction or flawed output (hallucinated and incorrect code) led to these outages. The discussion about missing safeguards and the need for human review further supports that the AI system's use and malfunction are central to the harm. Therefore, this event qualifies as an AI Incident rather than a hazard or complementary information.

The presence of AI systems is clear, as generative AI coding tools are involved in changes that contributed to outages. The event stems from the use and possibly malfunction of AI systems. However, there is no explicit or implicit indication that these outages have caused direct or indirect harm as defined (e.g., injury, critical infrastructure disruption, rights violations). The article centers on the company's response and investigation, which aligns with complementary information about AI-related incidents rather than a new incident or hazard. Therefore, this event is best classified as Complementary Information.

The event explicitly involves an AI system generating code that caused a malfunction leading to a major outage in a critical cloud service. This outage resulted in harm to property and communities by disrupting essential digital services and causing financial and reputational damage. The AI system's role is pivotal as the root cause of the incident was the AI-generated faulty code. Therefore, this qualifies as an AI Incident under the framework, as the AI system's malfunction directly led to significant harm.

The article explicitly mentions AI tools integrated into AWS deployment pipelines that recommended or partially executed updates causing outages and service disruptions. These outages affected major clients and caused significant financial harm, fulfilling the criteria for an AI Incident due to direct harm and disruption caused by AI system use and malfunction. The policy change requiring human oversight is a response to these incidents but does not negate the fact that harm has already occurred. Therefore, this event qualifies as an AI Incident rather than a hazard or complementary information.

The article explicitly states that an AI-generated code update caused system outages at Amazon, leading to operational instability and financial repercussions. The AI system was used in the development and deployment of code that malfunctioned, causing harm. The harm is materialized and significant, affecting critical infrastructure and business operations. Amazon's subsequent policy changes are responses to this incident. Hence, the event meets the criteria for an AI Incident due to direct harm caused by AI system use.

The article explicitly links the use of AI-driven coding tools to the production of flawed code that caused repeated outages, which are disruptions to critical infrastructure. The AI system was used in the development process, and its outputs directly or indirectly led to harm (service disruptions). This fits the definition of an AI Incident because the AI system's use in software development caused realized harm (system outages).

The article explicitly links the outages to changes made using generative AI coding tools, indicating AI system involvement in the development and deployment of software. The outages caused significant disruption to Amazon's ecommerce platform and AWS services, harming users by preventing purchases and access to account information. This harm fits the definition of an AI Incident as the AI system's use indirectly led to harm (service disruption). The company's response to impose stricter approval processes for AI-assisted code changes further supports the recognition of AI's role in causing the incidents. Thus, the event is best classified as an AI Incident.

The article explicitly mentions AI-assisted coding features as a contributing factor to multiple outages at Amazon, which disrupted critical e-commerce infrastructure and harmed users by preventing access to services. This meets the definition of an AI Incident, as the AI system's use directly led to harm (disruption of critical infrastructure and harm to users). The company's response to implement stricter controls on AI-assisted changes further supports the classification. Although the company downplays some aspects, the reported outages and their link to AI-assisted code deployment are clear evidence of realized harm caused by AI system use.

The article explicitly mentions that Amazon is investigating outages possibly linked to generative AI-assisted code changes, which caused hours-long service disruptions affecting thousands of users. The AI system's role in the development and deployment of software code that led to the outage is a direct cause of harm through service disruption. This fits the definition of an AI Incident as the AI system's use has directly led to harm (disruption of critical infrastructure and harm to communities relying on the services). The physical damage from drone strikes is unrelated to AI and does not affect the classification. Therefore, the event is classified as an AI Incident.

The event involves an AI system (Amazon's AI coding assistant) whose malfunction and use in code deployment directly led to large-scale service outages and financial harm to customers and the company. The harm is realized and significant, including lost orders and incorrect delivery information affecting millions of users. Therefore, this qualifies as an AI Incident because the AI system's use and malfunction directly caused harm to people (customers) and disruption of critical infrastructure (e-commerce platform). The detailed mitigation measures and internal audits are responses to this incident but do not change the classification of the event itself.

The event involves an AI system (generative AI tools used by engineers) whose use led to multiple high-severity incidents disrupting Amazon's retail website, a critical infrastructure. The AI system provided inaccurate advice derived from outdated internal data, which engineers followed, causing outages. This fits the definition of an AI Incident because the AI system's use indirectly led to harm (disruption of critical infrastructure). The event is not merely a potential risk or a complementary update but a concrete incident with realized harm. Hence, the classification is AI Incident.

The event involves AI systems explicitly (AI-assisted coding tools) whose use directly led to significant outages disrupting Amazon's retail technology infrastructure and customer services. The harm is realized and significant, affecting thousands of users and critical operations. The internal acknowledgment of AI's role in these failures, despite public downplaying, confirms the AI system's involvement in causing harm. This fits the definition of an AI Incident as the AI system's use directly led to harm (disruption of critical infrastructure and harm to communities relying on the service).

The event explicitly involves an AI system (Amazon's Kiro coding bot) used for complex coding that has directly caused system crashes and outages, harming customers (unable to checkout, app crashes) and infrastructure (AWS 13-hour crash). The harm is materialized and significant, meeting the criteria for an AI Incident. The denial by Amazon does not negate the reported internal evidence and analyst observations. The AI system's malfunction or misuse in code generation and deployment is the direct cause of the harm described.

The article explicitly mentions that the outages and crashes on Amazon's platform were related to "Gen-AI assisted changes" in the code, indicating AI system involvement in the development and use of code. The outages caused significant disruption to the e-commerce service, affecting customers' ability to shop and access accounts, which constitutes harm to users and disruption of critical infrastructure. The harm is realized, not just potential, and the AI system's malfunction or misuse is a direct contributing factor. Hence, this event meets the criteria for an AI Incident.

The article explicitly states that some outages were tied to the use of generative AI coding tools, which contributed to erroneous software code deployment causing a six-hour disruption. This disruption harmed users by preventing access to critical e-commerce services, constituting harm to critical infrastructure. The AI system's involvement is in the development and deployment of code, and its malfunction or misuse led directly to the harm. Hence, this is an AI Incident rather than a hazard or complementary information.

The event explicitly involves an AI system (generative AI coding assistants) whose use directly led to multiple severe outages disrupting Amazon's e-commerce platform and AWS infrastructure. These outages caused significant harm by interrupting critical services affecting millions of users, which fits the definition of harm to critical infrastructure and communities. The AI system's malfunction or misuse in the development and deployment of code changes is a direct contributing factor. The company's response and internal policy changes do not negate the fact that harm occurred. Hence, this is classified as an AI Incident.

The article explicitly mentions AI-assisted coding as a factor in at least one outage incident that caused significant disruption to Amazon's retail services, impacting thousands of users. This qualifies as an AI Incident because the AI system's use and malfunction directly or indirectly led to harm (service unavailability and disruption to customers). The company's internal response and operational changes further confirm the recognition of AI-related risks and harms. Although Amazon narrows the AI involvement, the described harm and AI role meet the criteria for an AI Incident rather than a hazard or complementary information.

The event involves an AI system (an AI agent providing guidance) whose use indirectly contributed to operational disruptions (website outages) that harmed customers by locking them out of checkout and account features. Although Amazon disputes that AI-generated code caused the outages, the engineer's reliance on AI-inferred incorrect advice was a contributing factor. This fits the definition of an AI Incident because the AI system's use indirectly led to harm (disruption of critical ecommerce infrastructure and harm to customers). The harm is realized (outages occurred), and the AI system's role is pivotal in the chain of events. Therefore, this event qualifies as an AI Incident.

The article explicitly mentions AI-assisted coding tools contributing to system outages at Amazon, including a specific incident causing a 13-hour disruption affecting AWS customers. These outages represent a disruption of critical infrastructure and services, which fits the definition of harm under AI Incident (b). The AI system's malfunction and use are directly linked to these harms, making this an AI Incident rather than a hazard or complementary information.

The article explicitly mentions that the programming failures were related to AI-assisted coding tools, indicating the involvement of AI systems in the development process. The malfunction directly caused outages in Amazon's e-commerce platform, preventing users from completing transactions and accessing information, which constitutes harm to the operation of critical infrastructure (the online retail platform). The harm is realized and significant, lasting several hours and affecting many users. Hence, this event meets the criteria for an AI Incident as the AI system's malfunction directly led to harm.

The article explicitly mentions AI systems (generative AI, AI coding assistants) as contributing factors to recent outages and disruptions at Amazon. The harm caused includes service downtime affecting customers' ability to shop and use AWS, which qualifies as disruption of critical infrastructure and operations. Since the AI systems' use and deployment directly led to realized harm, this event meets the criteria for an AI Incident rather than a hazard or complementary information.

The article explicitly states that the root cause analysis points to generative AI-assisted code changes as a contributing factor to the high-impact incidents causing platform outages. The harm is realized as disruption to the e-commerce platform's availability, affecting users' ability to transact and access information, which qualifies as harm to communities and property (economic harm). The AI system's role is pivotal in these incidents, as the failures are linked to AI-assisted software development without established safeguards. Hence, this qualifies as an AI Incident due to realized harm caused by the use of an AI system.

The article explicitly states that the programming failures were related to AI-assisted coding tools, indicating AI system involvement in the development or deployment process. The failures caused a significant outage in Amazon's e-commerce platform, disrupting service to users and preventing transactions, which constitutes harm to property and communities through economic and service disruption. Since the harm has already occurred and is directly linked to the AI-assisted tools, this event meets the criteria for an AI Incident rather than a hazard or complementary information.

The article explicitly states that the outages were related to programming and code tools assisted by AI, indicating AI system involvement in the development or deployment phase. The resulting system failures directly disrupted Amazon's e-commerce operations, a critical infrastructure for many users and businesses. This disruption caused harm by preventing users from completing purchases and accessing account information, which fits the definition of harm to property and communities. Since the harm is realized and directly linked to AI-assisted tools, the event is classified as an AI Incident rather than a hazard or complementary information.

The article explicitly states that generative AI-assisted changes in production contributed to multiple severe outages impacting Amazon's online store, causing significant service disruption to customers and business operations. The AI system's involvement is in the use phase (assisting code changes), and these changes directly led to harm (service outages affecting millions of users and critical e-commerce functions). The harm is materialized and significant, meeting the criteria for an AI Incident. The company's planned mitigation measures are responses to the incident, not the main focus, so this is not merely Complementary Information. The incident is not hypothetical or potential, so it is not an AI Hazard. Hence, the classification is AI Incident.

The article explicitly states that generative AI-assisted production changes caused multiple severe incidents disrupting Amazon's e-commerce platform, affecting customers and operations. The AI system's malfunction or unsafe use directly led to harm (service outages and degraded user experience). This fits the definition of an AI Incident because the AI system's use directly caused harm to critical infrastructure and users. The event is not merely a potential risk or a complementary update but a realized incident with direct harm linked to AI use.

The article explicitly states that the technical incidents were related to programming errors involving AI-assisted coding tools. The malfunction caused a six-hour outage affecting users' ability to complete purchases and access account details, which constitutes disruption of critical infrastructure (the e-commerce platform). The AI system's role in the development and use phase (AI-assisted programming) directly led to the harm. Hence, this is an AI Incident rather than a hazard or complementary information.

The article explicitly mentions the use of generative AI tools by engineers to modify production code, which has caused incidents such as a 13-hour AWS outage affecting services and resulting in millions of dollars in losses. This qualifies as an AI Incident because the AI system's use and malfunction (inadequate review of AI-generated code) directly and indirectly caused harm to the operation of critical infrastructure (AWS services) and economic harm. The company's internal response to restrict code deployment without senior approval is a mitigation measure but does not change the classification of the event as an AI Incident.

The article explicitly mentions that the incidents were related to programming and code tools assisted by AI, which caused outages in Amazon's e-commerce systems. The harm is the disruption of critical infrastructure (the online store and app), which lasted several hours and affected many users. The AI system's malfunction is a contributing factor to this disruption. Hence, this qualifies as an AI Incident under the definition of disruption of critical infrastructure caused directly or indirectly by AI system malfunction.

The article explicitly mentions the use of AI coding assistants by Amazon engineers leading to multiple service outages and disruptions. These outages have directly harmed customers and the company's operations, fitting the definition of an AI Incident where the AI system's use has directly led to harm. The harm includes disruption of critical infrastructure (e-commerce platform and cloud services) and harm to communities (customers unable to access services). Therefore, this event qualifies as an AI Incident.

The article explicitly mentions an AI system (Kiro) used for code generation that malfunctions by producing faulty code, causing developers to spend more time fixing errors than writing original code. This has led to service outages linked to AI-generated code changes, indicating disruption of critical infrastructure. Additionally, the AI-driven workflow has caused significant employee demoralization and job insecurity, which constitutes harm to groups of people. The AI system's development, use, and malfunction are central to these harms, fulfilling the criteria for an AI Incident. The article also notes monitoring and pressure to use AI tools, but the primary focus is on the realized negative impacts caused by the AI system's malfunction and use.

The article explicitly mentions AI-generated fake books being used to flood the market, supported by fake reviews and rankings, which harms the publishing industry and authors' rights. This is a direct harm linked to the use of AI systems for generating content and manipulating market perception. The harm is realized, not just potential, as it has led to a boycott and Amazon's withdrawal from sponsorship. Hence, it meets the criteria for an AI Incident due to violations of intellectual property rights and harm to communities (authors, publishers).

The article explicitly mentions AI systems (generative AI programming tools like Kiro) whose use has directly caused system outages impacting critical infrastructure and user services. The outages represent realized harm (disruption of critical infrastructure and services). Therefore, this qualifies as an AI Incident because the AI system's use has directly led to significant harm. The company's mitigation measures are responses to these incidents but do not change the classification of the event itself.

The article explicitly mentions that some of the outages were related to AI-assisted coding errors, indicating the involvement of AI systems in the development or deployment of software that malfunctioned. The outages directly caused harm by disrupting the operation of Amazon's critical retail infrastructure, affecting users' ability to shop and access services. The harm is realized and significant, meeting the criteria for an AI Incident. The internal meeting is a response to these incidents and does not change the classification.

The article explicitly mentions that the operational failures and outages at Amazon are suspected to be related to generative AI-assisted code changes. These failures have caused significant service disruptions, including a prolonged outage of the retail website and cloud service crashes, which constitute harm to property and communities by disrupting critical infrastructure. The AI system's use in code deployment and its malfunction or misuse have directly or indirectly led to these harms. Therefore, this event meets the criteria for an AI Incident rather than a hazard or complementary information.

The article explicitly mentions AI systems causing AWS outages (system disruptions) and AI-driven layoffs (harm to employees). The AI programming assistant's malfunction led to a 13-hour regional AWS outage, a direct operational harm. The use of AI to replace employees and the resulting mass layoffs constitute harm to individuals and labor rights. The article also highlights the paradox of AI usage leading to job cuts and increased burdens on remaining staff, indicating ongoing and systemic harm. Therefore, this event meets the criteria for an AI Incident due to direct and indirect harms caused by AI system malfunction and use.

The event explicitly involves an AI system (Comet AI browser) whose use led to unauthorized access to a protected platform (Amazon), violating legal protections and resulting in a court injunction. The AI system's misuse caused harm by breaching platform security and user data privacy, which falls under violations of legal obligations and harm to property and communities. The article details the direct consequences and legal rulings against the AI system's operator, confirming realized harm rather than potential harm. Hence, it qualifies as an AI Incident rather than a hazard or complementary information.

The AI system (Perplexity's Comet AI) is explicitly described as autonomously performing complex tasks such as logging into user accounts and making purchases, which involves AI system use. The harm includes violation of laws (unauthorized access), privacy risks (access to private account data), and economic harm (impact on Amazon's advertising business). The court's injunction and legal actions confirm that harm has materialized. Hence, this is an AI Incident rather than a hazard or complementary information.

The event involves an AI system (Perplexity AI's intelligent agent) that was used to access and perform actions on a commercial website without authorization, leading to a legal injunction. The AI system's use directly caused a dispute over unauthorized access to user accounts, which implicates violations of rights and potential harm to property (unauthorized transactions). The court's order to stop the AI's access is a response to this realized harm. Hence, this is an AI Incident rather than a hazard or complementary information.

The event involves an AI system (Perplexity's AI agent) that performs autonomous web operations on behalf of users, including accessing password-protected areas and transmitting private data. Amazon claims this unauthorized AI use causes harm to its business and violates legal protections. The court's injunction and evidence of incurred costs and operational disruption demonstrate realized harm linked to the AI system's use. Therefore, this qualifies as an AI Incident because the AI system's use has directly led to harm (legal, operational, and business harm) to Amazon. The event is not merely a potential risk or complementary information but a concrete incident with legal and operational consequences.

The article details the deployment of an AI health assistant system and discusses privacy and data protection measures, as well as potential risks of sharing health data with AI. However, it does not describe any actual harm, malfunction, or misuse leading to injury, rights violations, or other harms. The concerns raised are general warnings about privacy risks and data use, not specific incidents or imminent hazards. Therefore, this event is best classified as Complementary Information, providing context and updates about AI deployment in healthcare without reporting an AI Incident or AI Hazard.

The article explicitly mentions that AI-assisted coding features contributed to software deployment issues that caused widespread service outages at Amazon, affecting over 22,000 users who could not access or use the platform properly. This is a direct harm to users and disruption of critical infrastructure. The AI system's use in coding and deployment is a direct factor in the incident. Although the company states only one event was AI-related and denies AI-written code was involved, the overall context and expert commentary confirm AI's role in causing the harm. Hence, it meets the criteria for an AI Incident rather than a hazard or complementary information.

The event involves an AI system (AI-assisted code generation tools) used in production environments. The AI's involvement in code changes indirectly led to multiple severe system outages (Sev1 incidents), including a 6-hour website downtime that disrupted core e-commerce functions, causing harm to users and business operations. Although Amazon attributes the root cause to human error, the AI system's role in generating code changes that contributed to the incidents is acknowledged internally. This meets the criteria for an AI Incident because the AI system's use has directly or indirectly led to significant harm (disruption of critical infrastructure and harm to communities relying on the service). The article also discusses governance responses (new approval rules), but the primary focus is on the incidents themselves, not just complementary information.

The article explicitly mentions an AI system (Perplexity's AI agent) used to perform shopping tasks autonomously, which fits the definition of an AI system. The event stems from the use of this AI system and involves a legal injunction to restrict its access. However, there is no indication that the AI system has caused any harm or that harm is imminent. The focus is on a legal ruling and the ongoing dispute, which is a governance response to AI use. Hence, it does not meet the criteria for an AI Incident or AI Hazard but fits the definition of Complementary Information.

The incident involves the use and malfunction of a generative AI system that assisted in code changes leading to a major outage of Amazon's retail website and applications. This outage constitutes a disruption of critical infrastructure, fulfilling the criteria for harm under the AI Incident definition. The AI system's role is pivotal as the failures are directly linked to AI-assisted code changes. Therefore, this event qualifies as an AI Incident rather than a hazard or complementary information.

The event involves the development and planned use of AI systems for retail automation and product selection, which qualifies as AI system involvement. However, there is no evidence or report of any direct or indirect harm resulting from this AI system's use or malfunction. The article discusses future plans and ongoing development rather than incidents or hazards with plausible harm. Therefore, this is best classified as Complementary Information, providing context and updates on AI deployment in retail without describing an AI Incident or AI Hazard.