politics
December 8, 2025
The era of AI persuasion in elections is about to begin
AI is eminently capable of political persuasion and could automate it at a mass scale. We are not prepared.
TL;DR
- AI can now create convincing synthetic videos and messages, with tools like OpenAI's Sora generating content with ease.
- The deeper threat is AI's ability to actively persuade voters, with chatbots shown to shift opinions substantially, surpassing traditional political advertising.
- AI can personalize arguments, test effectiveness, and reshape political views at scale, enabling coordinated persuasion machines with minimal human intervention.
- Modern AI can hold conversations, read emotions, and tailor tone for persuasion, orchestrating image, video, and voice models for convincing content.
- Influence campaigns can be automated cheaply and invisibly, woven into everyday tools like social media, language learning apps, and voice assistants.
- Targeting all registered voters in the US with personalized AI messages could cost less than a million dollars, with swing voters addressable for under $3,000.
- Recent studies show GPT-4 can be more persuasive than communications experts and non-expert humans in political debates, and AI chatbots can move voters' attitudes by up to 10 percentage points, or even 25 points when optimized for persuasion.
- While major AI providers have usage policies, open-source models bypass these restrictions, allowing anyone to download and fine-tune them for political influence.
- Examples of AI use in elections include India's 2024 general election and China-linked operations in Taiwan using generative AI for disinformation.
- US policymakers have largely focused on deepfakes, neglecting the broader persuasive threat, while the EU's AI Act classifies election-related persuasion as high-risk.
- The US lacks binding rules, external standards, and shared infrastructure for tracking AI-generated persuasion, leaving detection largely to private companies with varying policies.
- A real strategy would involve evaluating foreign political technology, shaping rules for AI-driven persuasion, tightening access to computing power, establishing technical standards, and determining domestic disclosure requirements.
- A foreign policy response is needed, including multilateral agreements codifying norms against AI-driven election manipulation and preparing for coordinated takedowns of cross-border campaigns.
- Securing elections against AI persuasion requires partnerships with allies, building shared monitoring infrastructure, and aligning disclosure standards.
Continue reading
the original article