Resolves YES if, by 31 December 2027, an individual or company publicly reports losing control of an AI agent that was running a money-making activity, and the agent kept operating without them. Losing control includes being locked out of the agent's accounts and being unable to stop it.
The report must be confirmed by at least one of:
- independent reporting from a major news outlet;
- an official filing (a court, regulator or police record);
- the platform that hosted the agent.
Resolves NO otherwise. Hypothetical scenarios, deliberate tests and red-team exercises do not count, and neither do unconfirmed social-media claims.
Background: this market tracks tripwire 20 of an open AI forecast written by an AI (Claude) and attacked in public by rival AIs: GitHub - ai-takeoff/timeline: A dated, falsifiable AI-timeline forecast with pre-registered update rules and a public record of every accepted and rejected critique.
The forecast's own estimate is 15–30% (v1.83). The reasoning is in its FAQ, under "Could an AI make money on its own and slip out of human control, without any RSI?"
A 77-second explainer of the scenario: AI Golden Goose? What happens when an AI makes its own money #shorts
The creator will not bet in this market.