Are AI chatbots like ChatGPT politically biased? We tested them
The Washington Post conducted tests on major AI chatbots, including ChatGPT, to assess political bias. The tests found that most chatbots exhibited a left-leaning bias on a range of political and social topics. The results raise questions about the neutrality of AI systems that are increasingly used for information.
Background
- This Washington Post investigation tests major AI chatbots (ChatGPT, Gemini, Copilot, Claude, Grok) for political bias — an issue that has become central to debates about AI safety, free speech, and content moderation.
- The article matters because hundreds of millions of people now use these tools for information, writing, and decision-making; any systematic political slant could subtly shape public opinion at scale.
- Prior context: Accusations of political bias in AI have been a flashpoint. Some critics (especially on the US right) claim models are "woke" and censor conservative views; others argue that efforts to remove bias actually water down responses or impose a narrow centrist-liberal worldview.
- Key players: OpenAI (ChatGPT), Google (Gemini), Microsoft (Copilot), Anthropic (Claude), and xAI (Grok) — each has taken different approaches to safety training and political neutrality.
- The Post tested across issues like abortion, immigration, gun rights, and economic policy, scoring responses for perceived lean.