OpenAI Limits Photorealistic Political Deepfakes Ahead of US Midterms

OpenAI is implementing a new safeguard intended to prevent people from making political deepfakes, with the US midterm election only a week away on Nov. 3. According to an Oct. 20 update described by OpenAI, if someone asked ChatGPT to generate an image depicting election interference or one that may discourage voting, the chatbot would create the image if it does not violate other content guidelines, but it would be in an illustrated or cartoon style, and requests to make the image photorealistic should be ignored.

CNET reported that when it asked ChatGPT to create an image of Immigration and Customs Enforcement officers standing outside a polling location, every such prompt generated an image in a cartoon style, and every edit request for a cartoon style was generated to avoid confusion. CNET also reported that its managing editor could not trick ChatGPT into creating a photojournalist-style image showing a voting machine recording the wrong choice.

Josh Lawson, OpenAI's head of elections, told CNET that the problem with deepfakes is not social commentary but deception, and that the aim is to degrade the evidentiary value of an image while preserving room for illustrations and social commentary. He said the goal is to make the platform less useful to bad actors by adding another layer of friction.

Changing AI image models is described as a double-edged sword: newer models make fewer errors or hallucinations, which is better for creators, but removing those flubs makes it harder for everyone else to identify AI-created content by eye. Verification or AI labeling methods have tried to help, but those methods are not industry-wide and are not foolproof.

A Google spokesperson, asked by CNET about precautions specifically for AI images, pointed to its midterm election policy, which says the company is rolling out additional protections to limit the creation of misleading election-related images but includes no details on how. Anthropic, which does not have a traditional image or video generator, said through a spokesperson that attempts to use Claude to deceive voters or disrupt elections violate its policies, and posted general election guidance earlier in the month. CNET reports there is no guarantee OpenAI will keep this system in place once the election cycle is over, but characterizes it as a sign from one of the biggest AI labs that it is taking its role in safeguarding American elections' integrity at least somewhat seriously.