OpenAI and Anthropic set up Election Day teams to monitor AI misuse

In short: AI companies are running Election Day monitoring teams to spot and respond to attempts to use their tools to mislead voters.

What's going on

OpenAI and Anthropic have set up dedicated Election Day “war rooms,” which are staff teams that watch for problems in real time. Their goal is to detect misuse of AI tools during the U.S. midterm elections and respond quickly.

The main worry is that AI can make election misinformation faster to create and easier to spread. This includes convincing fake images and other content that looks real, like a forged document or a photo that never happened (think of a counterfeit bill, but for media).

OpenAI says its Election Operations Center reviews flagged activity, enforces company rules, and helps users find reliable voting information. It says it is working with Democracy Works for voting information and The Associated Press for election results. OpenAI also says its image safeguards may replace realistic images with cartoon-style images in cases where a realistic picture could mislead voters.

Anthropic says it has a dedicated team to handle misuse such as impersonating election officials or spreading false information. Its chatbot, Claude, directs users who ask about voting logistics to Democracy Works resources.

Other companies are making similar moves. Meta says it activated its Election Operations Center for real-time monitoring, and Microsoft says it refined Copilot’s election responses and points users to state and local election officials.

What to watch

These steps are not a guarantee that false content will not spread. Tests cited by the Brennan Center suggest protections vary by product, with ChatGPT improving against fake election images, while Google’s Gemini and Meta AI were reportedly more willing to generate convincing fakes than in an earlier test.

Source: NYTimes