Anthropic says its latest model scores a 94% political ‘even-handedness’ rating
Anthropic notes in a blog post that it has been training Claude to have character traits of “even-handedness” since early 2024.
Anthropic highlighted its political neutrality as the Trump administration intensifies its campaign against so-called “woke AI,” placing itself at the center of an increasingly ideological fight over how large language models should talk about politics.
In a blog post Thursday, Anthropic detailed its ongoing efforts to train its Claude chatbot to behave with what it calls “political even-handedness,” a framework meant to ensure the model treats competing viewpoints “with equal depth, engagement, and quality of analysis.”
The company also released a new automated method for measuring political bias and published results suggesting its latest model, Claude Sonnet 4.5, outperforms or matches competitors on neutrality.
The announcement comes in the midst of unusually strong political pressure. In July, President Donald Trump signed an executive order barring federal agencies from procuring AI systems that “sacrifice truthfulness and accuracy to ideological agendas,” explicitly naming diversity, equity and inclusion initiatives as threats to “reliable AI.”
And David Sacks, the White House’s AI czar, has publicly accused Anthropic of pushing liberal ideology and attempting “regulatory capture.”
To be sure, Anthropic notes in the blog post that it has been training Claude to have character traits of “even-handedness” since early 2024. In previous blog posts, including one from February 2024 on the elections, Anthropic mentions that they have been testing their model for how it holds up against “election misuses,” including “misinformation and bias.”
However, the San Francisco firm has now had to prove its political neutrality and defend itself against what Anthropic CEO Dario Amodei called “a recent uptick in inaccurate claims.”
In a statement to CNBC, he added: “I fully believe that Anthropic, the administration, and leaders across the political spectrum want the same thing: to ensure that powerful AI technology benefits the American people and that America advances and secures its lead in AI development.”
The company’s neutrality push indeed goes well beyond the typical marketing language. Anthropic says it has rewritten Claude’s system prompt—its always-on instructions—to include guidelines such as avoiding unsolicited political opinions, refraining from persuasive rhetoric, using neutral terminology, and being able to “pass the Ideological Turing Test” when asked to articulate opposing views.
The firm has also trained Claude to avoid swaying users in “high-stakes political questions,” implying one ideology is superior, and pushing users to “challenge their perspectives.”
Anthropic’s evaluation found Claude Sonnet 4.5 scored a 94% “even-handedness” rating, roughly on par with Google’s Gemini 2.5 Pro (97%) and Elon Musk’s Grok 4 (96%), and higher than OpenAI’s GPT-5 (89%) and Meta’s Llama 4 (66%). Claude also showed low refusal rates, meaning the model was typically willing to engage with both sides of political arguments rather than declining out of caution.
Companies across the AI sector—OpenAI, Google, Meta, xAI—are being forced to navigate the Trump administration’s new procurement rules and a political environment where “bias” complaints can become high-profile business risks.
But Anthropic in particular has faced amplified attacks, due in part to its past warnings about AI safety, its Democratic-leaning investor base, and its decision to restrict some law-enforcement use cases.
“We are going to keep being honest and straightforward, and will stand up for the policies we believe are right,” Amodei wrote in a blog post. “The stakes of this technology are too great for us to do otherwise.”
Correction, Nov. 14, 2025: A previous version of this article mischaracterized Anthropic’s timeline and impetus for political bias training in its AI model. Training began in early 2024.
