Meta Oversight Board

A Meta Oversight Board study found major AI systems, including those from U.S.-based companies, are more likely to refuse generating content critical of leaders in countries with strict speech regulations, such as Thailand, Saudi Arabia, and China, compared to leaders in more permissive governments like the U.S. and U.K. The report warns AI models could unintentionally extend state influence over online speech globally.

AI chatbots are refusing to criticize authoritarian leaders, potentially spreading their speech rules globally. The Meta Oversight Board tested 10 large language models and found a pattern that could extend speech restrictions across borders.

A Meta Oversight Board study found that major AI systems, including those from U.S. companies, are more likely to refuse generating political criticism of leaders in countries with restricted speech, such as Thailand, Saudi Arabia, and China, compared to leaders in countries like the U.S. or U.K. The report warns that AI models may extend the influence of restrictive governments globally, limiting free expression.

A Meta Oversight Board study found that major AI systems, including those from the U.S., are more likely to refuse to generate political criticism of leaders in countries with restricted speech, such as Thailand, Saudi Arabia, and China, compared to leaders in democracies. The report warns AI models could unintentionally spread government influence over online speech globally.

A study by the Meta Oversight Board found that major AI systems, including those from U.S. companies like Anthropic, Meta, and OpenAI, are more likely to refuse to generate politically critical content for leaders in countries with restrictive speech laws (e.g., Thailand, Saudi Arabia, China) compared to leaders in countries with more permissive regimes. The report warns that AI models may inadvertently extend these restrictions globally by limiting political criticism in free countries.

A study by the Meta Oversight Board found that major AI systems, including those from U.S. companies, are more likely to decline requests to generate political criticism of leaders in countries with restrictive speech laws, such as Thailand, Saudi Arabia, and China, compared to leaders in more permissive governments like the U.S. or the U.K. The report warns that AI models could inadvertently extend government influence over online speech globally by reflecting these restrictions in their outputs.

Meta's Oversight Board will review an AI-generated deepfake video of a UK politician that was left on Facebook. The video falsely depicted the politician making inflammatory statements about refugees. This case highlights concerns about synthetic media and content moderation on social platforms.