On May 27, 2026, OpenAI set out measures to support global elections and provided an update on October 20, 2026 ahead of U.S. midterms. The company said its priorities are surfacing reliable voting information, supporting cyber defenders, and increasing transparency about AI-generated content.
New tools and integrations
OpenAI announced a live integration with Democracy Works in the U.S., across languages, to provide registration details, deadlines, and voting locations inside ChatGPT. It also said an integration with The Associated Press will bring live vote counts into ChatGPT for elections in the U.S. and Brazil after polls close, while noting that election officials remain the authority on certified results.
To reduce the risk that generated images are mistaken for documentary evidence, OpenAI said it will apply image stylization for requests that depict election interference or could discourage voting, producing cartoonified illustrations rather than photorealistic images. The company emphasized that its policies against election interference and demobilization still apply to stylized outputs.
Support for defenders, provenance and policy
OpenAI described Daybreak as a set of programs to help harden software, including Codex Security, which the company said automatically identifies and helps remediate vulnerabilities in developers’ code. The Trusted Access for Cyber (TAC) program provides verified individuals access to frontier models for cyber defense. OpenAI said it has offered Codex Security and TAC access to voting-system manufacturers registered with the U.S. Election Assistance Commission and is engaging the National Association of Secretaries of State (NASS) and the National Association of State Election Directors (NASED).
On provenance, OpenAI announced deployment of SynthID watermarks on images generated through ChatGPT, Codex, or the OpenAI API, and continued use of the C2PA standard for metadata and cryptographic signatures. The company previewed a public verification tool to detect SynthID marks and surface C2PA metadata on images encountered off-platform.
OpenAI reiterated enforcement of Usage Policies that bar political ads on its platform and prohibit creating or distributing scaled campaign messaging for or against candidates, parties, or ballot measures, while allowing human-directed campaign tasks such as briefings, translation, and administrative work. The company also expressed support for legislation including Protect Elections from Deceptive AI Act (S. 1213) and AI Act (S. 2346).
OpenAI said it has published regular reports on misuse and continues political-bias evaluation work under its Model Spec principle Seeking the Truth Together, and that it will keep adapting safeguards in collaboration with partners and election authorities.
Original source: OpenAI News