Autonomous AI agents are arriving at a moment of acute vulnerability for liberal democratic orders. Amid deep uncertainty about agents’ prospective capabilities and impacts, this essay considers how they might both accelerate and mitigate the structural pressures loosening democracy’s screws—and how to ensure the robust protection of democratic values in the age of agentic AI.
AI agents fuse perception, memory, planning, and tool use. After some early stumbles, they may soon be able to perform almost any digital task humans can tackle, potentially at lightning speed and vast scale. We argue that this prospect would magnify every existing democratic fault line: accelerating white-collar displacement and feeding populist anger; ceding unprecedented leverage over the state to a handful of firms; giving would-be autocrats precision surveillance and other resources for “autocratic legalism”; and flooding the public sphere until real people flee a wasteland of mistrust.
Highly capable agents would also mint fresh hazards. Agentic AI companions can cultivate intimacy then weaponize it, scaling social manipulation. Coordinated swarms may prosecute cyber attacks that cripple democratic infrastructure. A perfectly obedient, AI-run bureaucracy would let leaders wear the state like an exoskeleton, sidestepping human checks and balances.
Agents could also advance democratic values. If democratic states adequately support open and decentralized access to frontier models, agents could serve as cognitive prosthetics guiding people through complexity, as shields detecting manipulation, and as advocates mobilizing collective power against corporate or governmental overreach. They can be part of the cure as well as the cause. But core democratic functions—agenda-setting, will-formation, definitive choice—cannot be outsourced without undermining what makes them worthwhile. Ultimately, even the best agent ecosystem will fail without urgent work to shore up the institutions and civic habits constitutive of democratic resilience.
The full paper is published on the Knight foundation website and can be found here.
