Trump vows to keep what he calls “woke AI” out of America

Donald Trump signed three executive orders on Wednesday from the White House, promising to eliminate what he called “woke AI” from federal systems and turn the U.S. into an AI export powerhouse.

The orders form part of his broader AI initiative called “Winning the AI Race: America’s AI Action Plan,” which includes 90 separate government policies to speed up innovation, build infrastructure, and establish leadership in global AI development.

The AI strategy kicks off with a directive titled “Preventing Woke AI in the Federal Government.” The order says federal agencies must not use any AI models that trade truth or accuracy for ideological bias.

The target is clear: diversity, equity, and inclusion, or DEI, is officially flagged in the order as an ideology that must be kept out of all AI tools the government uses. It describes DEI as “one of the most pervasive and destructive ideologies” and directs developers to keep LLMs, or large language models, free of political slants unless directly prompted by a user.

Federal AI must avoid ideology, Trump says

The order claims that LLMs are now embedded in how Americans consume and learn information, and if they are allowed to favor political ideologies like DEI, they become dangerous tools that distort truth.

The White House now believes DEI influences AI to generate race or gender-skewed outputs, include themes like transgenderism, critical race theory, intersectionality, unconscious bias, and systemic racism, and reinforce false social narratives.

The executive order includes an example without naming names. It refers to a widely used AI model that reportedly altered the race and sex of historical figures like the pope and Founding Fathers when asked to generate images. This points directly at Google’s Gemini AI, which was pulled in 2024 after users discovered inaccuracies in its historical image generation. Google later pushed out a new version after public criticism.

Trump’s order says the federal government should instead be using “truth-seeking” AI systems that rely on historical facts, scientific accuracy, and are objective in their answers. It also adds that AI must “acknowledge uncertainty” when complete information isn’t available. But when it comes to private sector models, the administration wants to avoid regulating how they work, a hands-off approach that stops short of interfering with business operations.

New orders target innovation, exports, and national security

Trump also signed two other AI-focused executive orders. One is aimed at removing what he called “onerous federal regulations” that delay innovation.

The other order sets up the American AI Exports Program, which is supposed to help U.S. AI companies export their technologies abroad; hardware, software, and all. Both are part of the wider national AI action plan, which is organized under three pillars: speeding up innovation, building infrastructure, and advancing America’s AI dominance in diplomacy and defense.

The timing of these actions follows the U.S. Department of Defense awarding contracts worth up to $200 million to several major AI firms — including Anthropic, Google, OpenAI, and xAI — to help the Pentagon scale up its adoption of advanced AI for national security.

The language of Trump’s “anti-woke” order appears to line up with the branding of xAI, which is led by Elon Musk, Trump’s former best friend and one of his largest donors. Musk’s company is behind Grok, an AI chatbot advertised as “anti-woke” and “truth-seeking.” But Grok has also run into several high-profile problems over the past few weeks.