Story

Researchers have discovered that manipulating sentence structure can bypass AI safety rules, allowing users to generate content that circumvents content moderation systems. This finding highlights the ongoing challenge of developing robust AI systems capable of detecting such evasive techniques.