Killswitch Protocols [pdf] (summerofprotocols.com)

Killswitch Protocols argues that intentional, engineered “death” mechanisms—killswitches—are a necessary feature of complex socio-technical systems, from nuclear SCRAM buttons to derelict MMOs and large online communities. The paper frames killswitches as an acknowledgement that systems can spiral into catastrophic states, and surveys a governance/design spectrum: fully centralized manual controls, automated trigger-based shutdowns, distributed or democratic mechanisms, and hybrid arrangements with independent vetoes or layered overrides. It emphasizes practical distinctions (e.g., automated failsafes are feasible in tractable domains like power grids or financial clearing, but human judgment is often required for social systems) and highlights risks such as irreversibility, adversarial capture, and perverse incentive effects created by the mere presence of a killswitch. For the AI/ML community the work is a compact primer on safety and governance tradeoffs. Key technical implications include choosing appropriate trigger logic, balancing latency and representativity (fast autocracy vs slower democratic processes), designing overrides and vetoes to limit capture, and considering how protocol awareness reshapes participant incentives and compliance. The authors stress that killswitches are coordination mechanisms, not panaceas: they constrain powerful actors and reshape system incentives, but introduce attack surfaces and hard tradeoffs between efficiency, safety, and legitimacy—issues central to deployment and alignment of large-scale AI systems.