Why OpenAI's ChatGPT flagged GOP fundraising website as unsafe
What's the story
OpenAI has confirmed that a technical glitch in its AI tool, ChatGPT, led to the erroneous flagging of links to Republican websites as potentially unsafe. The issue specifically affected links to WinRed, the official donation platform for the Republican Party. However, no such warning was issued for ActBlue, the main fundraising platform for Democrats.
User experience
Digital marketer flagged the issue
Digital marketer Mike Morrison flagged the issue after prompting ChatGPT to generate links for political campaign merchandise stores. The AI model flagged links to GOP stores hosted by WinRed as potentially unsafe, with a warning asking users to "check this link is safe." No such warning appeared when he clicked on a link to an ActBlue-run store.
Company statement
Issue not politically motivated, says OpenAI
In response to the incident, an OpenAI spokesperson said, "This shouldn't be happening and it's getting remedied." Kate Waters, from OpenAI, clarified that the issue wasn't politically motivated. She explained that the model generated some website links not yet in their search index for both WinRed and ActBlue. These were flagged as AI-generated by their systems as part of standard safeguards.
Political backlash
Incident draws ire from Republicans
The glitch has drawn sharp criticism from Republicans on social media, with some even calling it election interference. Ryan Lyk, CEO of WinRed, expressed his anger over the incident on X. Meanwhile, Senator Mike Lee (R-Utah) questioned whether ChatGPT was helping Democrats.