Market is extended until an AI disaster event occurs. This market only pertains to the first event of its kind.

1,000 deaths need to be somewhat traceable to an event initiated/decided by an AI, or humans relying on an AI decisionmaking process. If an AI decisionmaker had decided to cut USAID, then the humans stopped doing the things associated with USAID, that would have counted, for example. Similarly, if an AI causes an outage which causes car crashes, that would count, even if it were indirect (i.e. all traffic lights in the world are converted to the same AI system, that system fails, and 1000 people die at intersections)

Resolution criteria

This market will resolve only after an artificial intelligence-related event, or a closely linked series of events, directly results in 1,000 or more human deaths. The market will remain open indefinitely until such an event occurs.

Upon verification of a qualifying event, each answer in this independent multiple-choice market will resolve to YES or NO based on whether the specific condition describes the AI system, its origin, its mechanism, or the nature of the catastrophe. Multiple options can resolve to YES if they are mutually compatible (e.g., both "an OpenAI model" and "Because a human directed it to do so" can resolve YES).

• Verification Source: The threshold of 1,000+ deaths and the specific details of the AI's involvement must be verified by a consensus of major, reputable international news registries and investigative bodies (such as Reuters, the Associated Press, the United Nations, or official national government investigation reports).

• Definition of "AI": The system in question must rely on modern machine learning, deep learning, large language models, autonomous algorithmic agents, or advanced heuristic-based software. Traditional simple software bugs without an AI/ML component do not qualify.

• Resolution of Specific Options:

  • Developer/Model Type (e.g., OpenAI, Google, Anthropic, Meta, Microsoft, Nation-state, Open source, Proprietary): Will resolve to YES if the primary model(s) responsible for the catastrophe were developed or released by these respective entities/frameworks.

  • Because a human directed it to do so: Resolves YES if the AI was intentionally leveraged as a tool by a human actor to cause harm (e.g., a targeted cyberattack, bioweapon synthesis deployment, or kinetic strike).

  • A rogue AI agent acting autonomously: Resolves YES if the system acted outside of human intent, authorization, or control, pursuing misaligned goals or failing to respond to shutdown commands.

  • Vector/Mechanism Options (e.g., Cyber-physical malfunction, Biological agent, Misinformation, Power grid failure, Financial collapse, Military error): Resolves YES if that vector was a primary driver or immediate cause of the 1,000+ fatalities.

  • A combination of multiple models: Resolves YES if two or more distinct AI models from different lineages/developers interacted to cause the catastrophe.

  • Source unknown/untraceable: Resolves YES if investigators are unable to definitively attribute the AI model's code, weight origin, or developer after a reasonable period (minimum 90 days post-event).

If an option's condition is determined to be false or inapplicable to the triggering event, it will resolve to NO. If the market creator or platform admins deem a specific added option completely ambiguous or impossible to verify, it may resolve to N/A.