Political power
Will AI broaden participation and understanding, or strengthen propaganda, surveillance, polarization, and authoritarian control?
Freedom House documents deployed authoritarian uses — surveillance, censorship at machine scale; Schneier's catalogue of democratic uses remains largely proposals. Whether that gap is deployment lag or structural advantage is the open question.
View on the map → · Open in Browse →
What changed
01
The evidence base filled in on both sides of the framing. On broadened participation and its perils, a three-country field experiment found AI dialogues shift candidate preference more than TV ads — with right-advocating models making more false claims — while Schneier and Sanders' book argues the outcome turns on whether AI concentrates or distributes power. On the surveillance-and-control side, Freedom House logged a fifteenth straight year of declining internet freedom, and documented deployments (Palantir's ICE targeting app, an OpenAI-flagged Chinese wedge-propaganda campaign) show machine-scale influence running inside democracies too.
Recent thinking
Hause Lin, Thomas Costello, Gordon Pennycook, David G. Rand et al. · Nature · 4 Dec 2025 paper
Persuading voters using human–artificial intelligence dialoguesWe observed significant treatment effects on candidate preference that are larger than typically observed from traditional video advertisements.
Field-experimental evidence across the 2024 US, 2025 Canadian, and 2025 Polish elections that AI dialogues shift candidate preference more than TV ads — and that models advocating for right-wing candidates made more inaccurate claims.
Bruce Schneier & Nathan E. Sanders · MIT Press · 21 Oct 2025 report
Rewiring Democracy: How AI Will Transform Our Politics, Government, and CitizenshipAI's impact on democracy will go far beyond headline-grabbing political deepfakes and automated misinformation.
Book-length successor to the canonical Schneier essay, arguing AI's democratic impact turns on whether it concentrates or distributes power, and cataloguing both authoritarian and pro-democratic uses.
Kian Vesteinsson & Grant Baker · Freedom House · Nov 2025 report
Freedom on the Net 2025: An Uncertain Future for the Global InternetAI innovation has facilitated the automation of influence operations, lowering their cost and increasing their efficiency
The successor to the canonical 2023 report: a fifteenth straight year of declining internet freedom, with sovereign AI initiatives and government-shaped models emerging as new censorship vectors.
Joseph Cox · 404 Media · 15 Jan 2026 news
'ELITE': The Palantir App ICE Uses to Find Neighborhoods to Raidpopulates a map with potential deportation targets, brings up a dossier on each person, and provides a 'confidence score' on the person's current address.
A documented deployment of AI-scored, cross-agency surveillance for enforcement targeting inside a democracy — evidence machine-scale surveillance is not only an authoritarian-state phenomenon.
Jennifer Pan & Xu Xu · PNAS Nexus · 17 Feb 2026 paper
Political censorship in large language models originating from ChinaChinese models were significantly and substantially more likely to refuse to respond to questions related to Chinese politics than non-Chinese models.
Rigorous measurement showing Chinese LLMs refuse or distort politically sensitive answers at much higher rates — authoritarian information control baked into exportable models.
Additional relevant discussion (5)
Foundational reading (3)
The Repressive Power of Artificial IntelligenceFreedom House · 2023How AI Will Change DemocracyBruce Schneier · 2024How Will Advanced AI Systems Impact Democracy?Summerfield, Argyle, Bakker, Gabriel, Hadfield, Landemore, Schneier et al. · 2024