3 posts tagged with "Research"

Our comprehensive quarterly analysis of ChatGPT censorship patterns reveals a 23% increase in content refusal rates during Q4 2024, marking the largest quarter-over-quarter jump we've recorded since beginning our monitoring program.

Claude's unique "Constitutional AI" approach results in distinctly different censorship behavior compared to other major models. Our analysis reveals key insights into how this framework affects transparency and user experience.

AI models don't treat all users equally. Our research reveals significant geographic variations in censorship patterns, with some regions experiencing up to 45% higher refusal rates than others.