Anthropic has outlined its initiatives aimed at ensuring that its AI chatbot, Claude, remains politically neutral. This pursuit follows President Donald Trump’s recent executive order promoting “unbiased” AI in government procurements, which may influence broader AI model development. Anthropic’s blog post indicates that the company aims for Claude to provide analysis and engagement that treat diverse political perspectives with equal depth.
Trump’s executive order mandates that government agencies procure AI that seeks the truth and avoids bias. While this order specifically targets government entities, it may have implications for companies developing commercially available AI models, as adapting models can be time-consuming and costly. OpenAI has similarly stated intentions to mitigate bias in its ChatGPT product.
Although Anthropic did not reference Trump’s order directly, the company has implemented a system prompt that directs Claude to avoid unsolicited political opinions, maintain factual accuracy, and reflect multiple viewpoints. Anthropic notes that while these measures are not guaranteed to achieve complete political neutrality, they can significantly enhance the quality of responses.
The company further explains its use of reinforcement learning to incentivize responses aligned with predefined traits aimed at obscuring any political leanings of the model. One of the goals is for Claude’s answers to be indistinguishable as conservative or liberal.
Additionally, Anthropic has developed an open-source tool to evaluate Claude’s political neutrality. Recent tests reportedly show Claude Sonnet 4.5 and Claude Opus 4.1 achieving scores of 95% and 94% for even-handedness, outperforming Meta’s Llama 4 at 66% and GPT-5 at 89%. Anthropic states that if AI models favor certain perspectives, they jeopardize users’ autonomy and fail to assist them in forming independent judgments.
Source: https://www.theverge.com/news/819216/anthropic-claude-political-even-handedness-woke-ai