The US midterms are getting closer, and OpenAI wants to be ready this time. In a series of announcements, the company laid out new measures against election manipulation and AI-generated disinformation.
Codex Security for election officials
OpenAI is offering its cybersecurity products — specifically Codex Security and the Trusted Access for Cyber program — to registered voting system manufacturers in the US. The company has also briefed the National Association of Secretaries of State and the National Association of State Election Directors on current cyber threats.
It’s an interesting move. OpenAI is positioning itself not just as an AI provider, but as an active partner in protecting democratic infrastructure. Whether election officials will actually take them up on the offer remains to be seen.
SynthID against deepfakes
To combat manipulated images, OpenAI is deploying SynthID — digital watermarks embedded in AI-generated images. The watermarks are designed to remain detectable even after screenshots or image editing. SynthID is being used across ChatGPT, Codex, and the OpenAI API.
This matters because deepfakes have become a real problem in election campaigns. The watermarks don’t solve the problem completely — they only mark images created through OpenAI’s own tools — but they’re a start.
Verified election info in ChatGPT
OpenAI announced a partnership with Democracy Works. When US users ask questions about elections — where to vote, how to register — ChatGPT will display verified information. Starting this fall, live election results from the Associated Press will also be available directly in ChatGPT.
The bigger picture
These measures are necessary, and OpenAI deserves credit for being proactive. At the same time, it’s also self-interest: if AI-generated disinformation influences an election and ChatGPT plays a role, OpenAI would face a massive reputation problem — especially now, with the planned IPO this fall.
What’s missing is an industry-wide solution. OpenAI can secure its own tools, but the AI landscape is fragmented. Open-source models that operate without watermarks, and smaller providers without comparable safeguards, remain a problem. This needs industry standards, not just individual initiatives.
Sources: