Key facts
- AI is being integrated into social media platforms for the first time in elections.
- AI can generate realistic images, video, and audio of candidates with low cost and speed.
- Generative AI chatbots can create fake news sites and influence campaigns at scale.
- AI tools like Google's Gemini, OpenAI's GPT-4, and Meta's Llama 2 have shown poor accuracy on election-related questions.
- Traditional AI could fuel mass challenges to voter eligibility, risking wrongful purges.
- Tech companies pledged to detect deceptive AI media and curb AI-fueled election trickery.
Artificial intelligence is increasingly embedded in election processes, presenting new dangers of disinformation and fraud. This year marks the first national American election where AI can generate convincing fake images, video, and audio of candidates at scale and low cost. Generative AI chatbots can also quickly create fake news sites and enable large-scale influence campaigns. AI tools have shown poor accuracy on election-related questions, and traditional AI could be used to challenge voter eligibility, potentially leading to wrongful purges. Tech companies, including OpenAI, Apple, Meta, Microsoft, TikTok, Google, and X, have pledged to take measures to detect and curb AI-fueled election trickery, such as helping audiences track AI-generated content and implementing "reasonable precautions." OpenAI has announced specific policies to prevent the misuse of its tools for impersonation and spreading voting misinformation, including encrypting images with digital codes. However, the effectiveness of these measures is uncertain due to their flexible language, and more robust action is deemed necessary.