Technology

OpenAI Fights Election Deepfakes With Cartoon Filter

OpenAI Fights Election Deepfakes With Cartoon Filter

OpenAI is implementing a new safeguard to prevent people from creating realistic political deepfakes during election season. The measure ensures that any user asking the chatbot for images depicting election interference or voting discouragement will receive an illustrated or cartoon-style picture instead of a photorealistic one.

The Cartoonification Safeguard

According to an October 20 update from OpenAI, ChatGPT will ignore requests for photorealistic political images if the prompt depicts election interference or anything that may discourage voting. Testing by CNET showed that prompts asking for Immigration and Customs Enforcement officers outside a polling location or a voting machine recording the wrong choice consistently generated cartoon-style graphics rather than realistic photos that could cause panic.

Josh Lawson, OpenAI’s head of elections, explained the philosophy behind the update to CNET. “The problem with deepfakes isn’t social commentary, it’s deception. So our aim is to degrade the evidentiary value of an image, while preserving room for illustrations and social commentary,” Lawson told CNET, noting that the goal is to make the platform less useful to bad actors by adding friction.

Industry Responses and Future Outlook

While advanced AI models produce better content for creators, the lack of flaws makes identifying AI-generated content much harder for the general public, as verification and labeling methods are not yet industry-wide or foolproof. Other tech companies are taking different approaches to address election integrity.

A Google spokesperson pointed to the company’s midterm election policy regarding its Gemini maker, stating that additional protections are rolling out to limit misleading election-related images without providing specific details. Meanwhile, Anthropic relies on policy enforcement, with a spokesperson telling CNET that any attempts to use Claude to deceive voters or disrupt elections violate company rules.

It remains unconfirmed whether OpenAI will maintain this cartoon-conversion system once the current election cycle concludes.

You May Also Like

Leave a Comment

Your email address will not be published. Required fields are marked *