How OpenAI is Approaching 2024 Worldwide Elections

OpenAI is implementing multi-faceted safety measures throughout the 2024 worldwide election process to prevent potential misuse, increase transparency around AI-generated content, and provide accurate voting information.

To provide trustworthy information, OpenAI partnered with the National Association of Secretaries of State (NASS) to direct users asking about voting methods and locations to CanIVote.org. It also advised users to check credible news outlets such as Associated Press or Reuters for questions about election results.

For deepfake prevention, guardrails were applied to refuse generating images of real people, including politicians. During the month immediately before the election, OpenAI blocked more than approximately 250,000 requests to generate images of major politicians such as Trump, Harris, and Biden using DALL·E.

Blocking threat actors and monitoring are also being carried out in parallel. OpenAI is focused on identifying and blocking the use of AI tools in covert influence operations, and to date, no evidence has been found that influence operations using OpenAI models related to the U.S. election have gone viral or secured a sustained audience.

Additionally, OpenAI is strengthening transparency of AI content by providing researchers with tools to identify images generated by DALL·E 3, and by participating in C2PA, a digital content authentication standard.