Overview
- OpenAI created an Election Operations Center this week to spot misuse of ChatGPT and other models and to enforce policy through the Nov. 3 vote, a step the company says will help limit the spread of AI-driven misinformation.
- Independent trackers show hundreds of AI-made political pieces this cycle, with the Wesleyan Media Project reporting about 220 AI-generated ads and more than $137 million in spending as of early October.
- A university research team has cataloged roughly 482 deepfake incidents so far and found President Donald Trump was the subject of about 279 of those posts, underscoring concentrated targeting and wide use of synthetic likenesses.
- Detection and voter risk remain serious problems because most people struggle to spot audio or video deepfakes and tools lag behind generation methods, with tests showing single-digit detection rates for some audio fakes.
- States have moved faster than Washington on rules, with the NCSL counting 31 states that have passed deepfake-related laws, but disclosure requirements vary and platforms restricting researcher access make overall tracking incomplete and likely undercounts.