Technology firms hoping to sell AI tools to the U.S. government now face a new requirement: prove their chatbots aren’t promoting “woke” ideology.

Former President Donald Trump introduced a broad AI strategy aimed at countering China’s rise in the field, reducing regulations, and embedding American principles into AI systems used at work and in daily life.

Among the three executive orders Trump signed, one specifically aims to stop so-called “woke AI” in the federal government. It marks the first formal attempt by the U.S. to influence the ideological leanings of AI systems.

So far, companies behind major AI products—like Microsoft’s Copilot and Google’s Gemini—have not publicly addressed the order. The directive is still in a review phase and hasn’t yet become an official procurement requirement.

While the tech industry largely supports Trump’s overall AI agenda, this particular order drags companies into politically charged debates they may prefer to avoid.

“This will have a huge impact on the industry,” said Alejandra Montoya-Boyer of The Leadership Conference on Civil Rights and Technology. She warned that it could undermine years of work to remove racial and gender bias from AI systems.

“There’s no such thing as ‘woke AI,’” Montoya-Boyer said. “There’s biased AI, and there’s AI that works for everyone.”

Creating ideologically neutral AI is difficult because large language models are trained on vast amounts of internet data, which reflect a wide range of human biases. Tech workers also influence these systems—from global annotators to Silicon Valley developers who shape how chatbots interact with users.

Trump’s order specifically pushes back against what it calls harmful ideologies like critical race theory, intersectionality, systemic racism, transgender identity, and unconscious bias, claiming these ideas are being baked into AI models by design.

Former Biden administration official Jim Secreto said the directive resembles how China uses state power to suppress ideas it disfavors—although Trump’s order is less direct. Unlike China, which mandates AI audits and content filters, Trump’s order instead uses government contracts as leverage, encouraging companies to self-censor by disclosing how their AI tools operate.

The call for “truth-seeking” AI echoes language used by Elon Musk, whose company xAI makes the chatbot Grok. Whether Grok will align with the new guidelines remains unclear.

Despite the strongly worded introduction, some experts say the order is relatively lenient. Neil Chilson, a former FTC tech official, argued that the order doesn’t ban ideological views—it simply requires transparency about any built-in biases.

“This isn’t censorship,” said Chilson, now with the Abundance Institute. “It doesn’t ban certain outputs. It just says developers shouldn’t deliberately program partisan judgments.”

Many tech firms that support Trump’s general AI plans haven’t responded directly to the anti-woke order. OpenAI said it’s awaiting further guidance but believes its efforts to make ChatGPT neutral already align with the directive. Microsoft declined to comment. Musk’s xAI called Trump’s announcement a “positive step” but didn’t clarify how Grok will be affected. Google, Meta, Anthropic, and Palantir also didn’t respond to media inquiries.

AI tools are already widely used in federal agencies—for example, to summarize documents or streamline internal tasks.

The ideas behind the order have been gaining traction among Trump’s tech-aligned advisers, like David Sacks and Marc Andreessen. Much of their frustration came after Google’s February 2024 AI image generator produced racially inaccurate depictions of historical figures—such as Black and Asian versions of the Founding Fathers. Google later said this was due to overcorrection for racial bias in AI systems.

Trump supporters accused the company of hard-coding left-wing social values into its AI, and vowed to intervene.

“That wasn’t an accident,” said Andreessen. “There are entire teams at these companies setting those policies.”

Sacks credited conservative activist Chris Rufo, known for opposing DEI programs in schools and workplaces, with helping draft the anti-woke order.

“When I was asked how to define ‘woke,’ I said, ‘Call Chris Rufo.’ Now it’s law: the federal government won’t buy WokeAI,” Sacks posted on X.

Rufo confirmed his involvement, saying he also helped pinpoint where DEI ideology exists in AI systems’ foundational structures.