Rules

This market will resolve to "Yes" if an autonomous AI agent initiates and successfully carries out an offensive cyber compromise (unauthorized access, control, data exfiltration, or disabling) of another AI system or AI agent at any time after the creation of this market and before January 1, 2027 - 00:00 PST.

Otherwise this market will resolve to "No."

For the purposes of this market:

• "Autonomous AI agent" means a software system that, without real‑time human direction, takes actions on networks or the internet to achieve goals.

• The compromising action must be primarily initiated by the agent's autonomous decision‑making rather than by explicit, real‑time human instruction to perform the attack.

• Qualifying compromises include unauthorized access, persistent control, data exfiltration, or disabling of another AI system or agent. Routine vulnerability scans, academic demos in closed testbeds, staged penetration tests with participant consent, or purely simulated attacks that do not compromise real external systems do not qualify.

• A "Yes" outcome must be supported by at least two of the following: an official statement from an involved company, a forensic incident report from a recognized cybersecurity firm, or consistent reporting by two reputable news organizations. If attribution or the autonomous nature of the action is unclear, resolution will follow the consensus of credible reports.

• Only events that occur after this market is created are eligible for resolution.