MEASURING POLITICAL BIAS IN CLAUDE

Summary from the TLDR AI newsletter

Anthropic released an open-source "Paired Prompts" evaluation that measures political even-handedness across AI models.