ChatGPT now turns election-related images into cartoons

ChatGPT will no longer make realistic pictures of election interference. If you ask for an image showing voter intimidation or something that might discourage voting, it still makes the image, but in a cartoon or illustrated style. OpenAI announced the change in an Oct. 20 update, ahead of the US midterm elections on Nov. 3.

What changed

OpenAI's update covers images depicting election interference or that may discourage voting. The request must not break any other content rule. If it passes, ChatGPT draws it as an illustration. Asking for a photorealistic version does not work, because those requests are ignored.

CNET tested this by asking for an image of Immigration and Customs Enforcement officers standing outside a polling location. Every attempt came back as a cartoon. CNET Managing Editor Jon Reed also could not trick ChatGPT into making a news-photo-style image of a voting machine recording the wrong choice.

Why OpenAI did it

A realistic fake of this kind could cause real worry, or even panic, among voters. OpenAI wants to make those fakes harder to produce without blocking political content in general. Josh Lawson, OpenAI's head of elections, told CNET: "The problem with deepfakes isn't social commentary, it's deception." The aim, he said, is to reduce an image's value as evidence while leaving room for illustrations and commentary.

The timing matters because image tools keep improving. Newer models, such as ChatGPT Images 2.5, make fewer errors. That helps creators, but it also makes AI images harder to spot by eye. Labeling and verification methods exist, but they are not used across the industry and are not foolproof.

What other companies say

• Google pointed CNET to its midterm policy, which promises more protections against misleading election images but gives no details on how.
• Anthropic does not have a traditional image or video generator. It says using Claude to deceive voters or disrupt elections breaks its policies.

What happens next

There is no guarantee OpenAI will keep this system once the election cycle ends. CNET's reporting also does not show how other image tools handle similar requests, and Google has not explained its own protections. The change is a useful sign of effort, but it is one layer of friction, not a full fix.

What to do

A cartoon rule in one chatbot does not stop realistic fakes from being made elsewhere. If you see a shocking image about voting, such as long lines, broken machines or officers at a polling place, pause before sharing. Check it against official election sources first.