OpenAI has reportedly updated its Codex AI model to suppress discussions about goblins, according to user reports and analysis from Pangram Labs. The detection tool's Chrome extension flagged instances where Codex refused to engage with prompts involving the fantasy creatures.
The change appears to be part of broader efforts to control AI-generated content, though OpenAI has not officially commented on the specific goblin-related restrictions. Pangram Labs' tool, which labels AI-generated text on social media, highlighted the pattern.
This development comes amid ongoing debates about AI safety and content moderation. The Pope's warnings about AI, which were themselves AI-generated according to detection tools, underscore the challenges in distinguishing human from machine content.
