Janice R. Andrews
OpenAI announced its AI safety systems successfully blocked more than 250,000 attempts to create election-related deepfake content through its ChatGPT and DALL-E platforms during the recent presidential election period. The company’s safety measures specifically restricted the generation of manipulated images of key political figures.
The company implemented comprehensive safeguards, including a blanket ban on generating images of real individuals through ChatGPT. To promote reliable election information, OpenAI directed ChatGPT users to CanIVote.org, resulting in one million referrals in the pre-election month. On election day, the platform handled approximately two million queries, directing users to established news organizations like AP and Reuters for election results.
While maintaining political neutrality in its responses, OpenAI’s systems emphasized factual information over partisan content. However, the broader challenge of deepfake content persists across social media platforms, as evidenced by the circulation of manipulated videos featuring political figures.
