OpenAI opens an election operations center to watch ChatGPT misuse through the US midterms

With the US midterm elections set for November 3, OpenAI has set up a dedicated election operations center, Politico reported on October 28. The outlet describes it as a "war room" that brings together staff from several teams across the company, with headcount in the hundreds, and says it starts operating this week and will keep running until every vote has been counted.

What the center is meant to do

According to Politico, the center is led by Josh Lawson, OpenAI's head of civic and strategic risk, who described its purpose as "bringing our teams together to respond quickly to emerging issues." Its remit covers misinformation and other attempts to disrupt the vote, abuse of OpenAI products such as ChatGPT, reports of unexpected model behaviour, suspected breaches of the company's usage terms and real-time events. It will also keep an eye on two partner integrations: Democracy Works, which supplies voter registration and polling-place information, and the Associated Press, which provides live results.

The operations center itself is not mentioned on OpenAI's election safeguards page, so the staffing figure and the description of its scope come from Politico's reporting. Other platforms, Meta among them, have run comparable centers in past election cycles.

The safeguards already in place

In a post published on May 27, OpenAI said ChatGPT would display live AP vote counts on election night in the United States and Brazil, and would rely on Democracy Works for practical voting information. Images produced with ChatGPT, Codex or the API are to carry SynthID watermarks and C2PA metadata, and a public verification tool is in preview. The company says its usage policies bar election interference, voter demobilisation and scaled campaign messaging for or against candidates, and that its intelligence and investigations staff look for covert influence operations. It also described briefings for state election officials and backing for two Senate bills on deceptive AI in elections.

What remains uncertain

Politico notes that OpenAI reported earlier this month on so-called "false front" operations linked to Russia and Iran that used its models, a reminder that misuse is already happening. It also cites a Brennan Center study that found ChatGPT was the only one of six popular AI tools to make significant headway against election falsehoods since May. Whether a short-term monitoring effort can catch fast-moving deepfakes, many of which spread on other platforms, will only be clear after the vote. OpenAI has not said whether it will publish a report on what the center detects.