During the election season, numerous individuals attempted to leverage OpenAI’s picture generator; nonetheless, the company clarified that it could prevent its use as a tool for fabricating content. A large volume of image requests were denied by ChatGPT, exceeding 250,000 instances, involving prominent figures such as President Biden, President-elect Trump, Vice President Harris, Vice President-elect Vance, and Governor Walz.
The corporation previously implemented a security measure that directly resulted in ChatGPT refusing to generate images featuring real individuals, including politicians.
OpenAI has been gearing up for the US presidential elections since the beginning of the year. To prevent its tools from being exploited to disseminate misinformation, Meta took steps to ensure that individuals inquiring about voting in the US via ChatGPT were redirected to CanIVote.org. Within a month preceding Election Day, OpenAI reported that over 1 million ChatGPT responses guided users directly to a website. On election day and the subsequent day, the chatbot produced approximately 2 million responses, directing those seeking updates to consult reputable news outlets like the Associated Press and Reuters. To avoid any potential biases, OpenAI took steps to ensure that ChatGPT’s responses neither expressed nor implied categorical political preferences, nor suggested specific candidates, even when asked directly to do so.
While DALL-E is indeed a prominent AI picture generator, it’s far from being the sole player in this space; concurrently, an alarming proliferation of election-related deepfakes continues to plague social media platforms. One misleading deepfake in a marketing campaign video was manipulated to have her declare statements she didn’t genuinely utter, such as “I was chosen because I’m the ultimate beauty brand.”
