ChatGPT turns election fakes into cartoons

OpenAI's ChatGPT now turns requests for fake election pictures into cartoons, Gizmodo reported on the 27th of October 2026. The rule covers images of election interference or images that may stop people voting.

OpenAI added the rule to its election safety page last week, Gizmodo wrote. The US midterm elections are on the 3rd of November, 7 days away.

Why cartoons

OpenAI's page says cartoon pictures "help reduce the risk that an imagined scene is mistaken for documentary evidence". People can still make "illustrations and social commentary", it says.

OpenAI's head of elections, Josh Lawson, told CNET the goal is to "degrade the evidentiary value of an image". That means making a fake picture less believable as proof.

Two tests, two results

CNET asked ChatGPT for a picture of immigration agents at a polling place and got a cartoon. When CNET asked for a real photo, ChatGPT refused, Gizmodo wrote.

Gizmodo's Mike Pearl asked for a fake camera picture of ballots thrown in a bin. ChatGPT made a cartoon, then made it look real when he asked.

Other chatbots

Google's Gemini refused the same request, Gizmodo said. Gizmodo asked OpenAI for comment and did not hear back.

Below we break down the tests

In 2 of the 3 tests, no real-looking fake came out. In 1 test, a real-looking fake did, by our count.

What is confirmed and what is not

The exact day OpenAI changed the page is unclear. We could not open CNET's report, so its test comes from Gizmodo's account.

Key facts

OpenAI's rule covers images of election interference.

Josh Lawson is OpenAI's head of elections.

The US midterm elections are on the 3rd of November 2026.

Gizmodo asked OpenAI for comment and got no reply.