Safety3h ago
Chatbots Favor Left-Wing Misinformation, Study Finds
ChatGPT, Gemini, and Claude more frequently repeated political falsehoods from the left, while Grok skewed right. This study highlights persistent alignment biases across leading LLMs. Developers must now address these asymmetric vulnerabilities to prevent systemic misinformation. The findings suggest that current safety guardrails fail to neutralize political leanings during factual retrieval.