Openai Launching Election War Room To Track Interference Efforts

OpenAI is standing up an election “war room” ahead of the midterms, the company exclusively told Decoded by POLITICO.

The Election Operations Center will begin running this week until all votes are counted, to identify the potential spread of misinformation or other efforts to disrupt the elections. It’s meant in large part to combat abuse of the company’s own tools like ChatGPT, monitoring real-time events and fielding reports of unexpected behavior from models or cases of suspected violations of their platform’s terms.

“We’ve put considerable effort [into] ensuring we help users get reliable voting information, combat misuse of our tools and support election defenders,” Josh Lawson, OpenAI’s head of civic and strategic risk who is leading the center, told Decoded by POLITICO. “The Elections Operations Center builds on that yearslong work, bringing our teams together to respond quickly to emerging issues and enforce our policies as people head to the polls.”

OpenAI’s move comes amid broader concerns about AI’s potential to distort next week’s midterm elections through the generation of misleading images, audio and video as well as the spread of misinformation. The leading AI company is attempting to position itself as a defender of the democratic process, even as bad actors attempt to use AI tools from OpenAI and others tools to disrupt free and fair elections.

OpenAI has also found itself at the heart of a contentious midterm cycle, where anti-AI messaging has become a rallying cry for some campaigns. And as their AI tools integrate further into Americans’ lives, the company will only be under more pressure to prevent misuse and deliver accurate, timely and useful information about elections to their users. The war room is an effort to prove that they can successfully do so.

It’s also a novel move from a leading AI lab, though other tech companies have previously set up similar centers, including Meta.

Baked into OpenAI’s models are safeguards that attempt to stop malicious actors from creating misleading content. This includes a process of “cartoonification” — for example, if a user asks OpenAI to generate an image of a long line at a polling place that doesn’t actually exist, it will come back with a cartoon image rather than a photorealistic one.

A report from the Brennan Center published last week found that of the six most popular AI tools, ChatGPT was the only one that had made significant progress against election disinformation since May of this year, highlighting the success of its cartoonification efforts.

OpenAI’s new center — which will include multiple teams from across the company, with staffing in the hundreds — will actively review cases flagged for potential abuse. Earlier this month, OpenAI reported two “false front” operations run by Russia and Iran that used their models to “launder geopolitical, conflict-related messaging into their target audiences.”

The midterms are happening at a momentous time for AI frontier labs. Since 2024, AI models have vastly improved but have also broken out of company control multiple times. As the tools have become more advanced, those looking to disrupt elections can also use other AI systems in attempts to bypass safeguards.

In May, OpenAI announced a partnership with the nonpartisan organization Democracy Works to give voters reliable information about registration and polling locations. The company also announced an agreement with The Associated Press in order to provide up-to-the-minute election results. Both of these on-platform integrations will be monitored by the election war room.

A version of this story first appeared in Decoded by POLITICO, the must-read briefing on the people, power moves, policy battles and market shifts defining what’s next in global technology. Sign up to subscribe to the newsletter and listen to the podcast.