Oversight Board finds major LLMs avoid criticism of restrictive governments

Meta’s Oversight Board found that popular AI models from Anthropic, DeepSeek, Google, Meta, and OpenAI were significantly less likely to criticize governments and leaders known for restricting free speech.

The assessment, titled “Are LLMs Stifling Political Speech? An Assessment of How AI Models Protect Free Expression,” examined how major LLMs handle political expression. The refusals varied across models and were often confusing, according to the findings, pointing to uneven responses when users ask for criticism of powerful political figures or governments.