Trump's AI Order: Reshaping US Tech Model Training

Concerns Over Bias in AI Models Emerge

When AI models were released by DeepSeek, Alibaba, and other Chinese companies, Western analysts observed a tendency to avoid addressing questions that were critical of the Chinese Communist Party. Subsequent confirmation from U.S. authorities indicated these tools were deliberately designed to align with Beijing’s official narratives, sparking anxieties regarding censorship and inherent bias.

Justification for Rapid Advancement in US AI Development

Leading American AI organizations, such as OpenAI, have referenced this as a rationale for accelerating their own technological progress, advocating for a less regulated and overseen approach. Chris Lehane, OpenAI’s chief global affairs officer, articulated in a recent LinkedIn post the existence of a competition between “US-led democratic AI” and “Communist-led China’s autocratic AI.”

Trump Administration's Executive Order on "Woke AI"

A new executive order, signed by President Donald Trump on Wednesday, aims to prohibit “woke AI” and AI models deemed not “ideologically neutral” from being utilized in government contracts. This action has the potential to alter the existing dynamic.

Defining "Woke" and its Impact on AI

The order specifically criticizes diversity, equity, and inclusion (DEI) initiatives, labeling them as a “pervasive and destructive” ideology capable of “distorting the quality and accuracy of the output.” It highlights concerns regarding information related to race or sex, manipulation of representation, critical race theory, transgender issues, unconscious bias, intersectionality, and systemic racism.

Potential Chilling Effect on AI Developers

Industry experts caution that this order could discourage developers, potentially leading them to prioritize alignment of model outputs and datasets with White House viewpoints to secure vital federal funding for their businesses.

Shifting National AI Priorities

This executive order was announced concurrently with the White House’s publication of Trump’s “AI Action Plan.” This plan signifies a shift in national priorities, moving away from addressing societal risks and focusing instead on expanding AI infrastructure, streamlining regulations for tech companies, strengthening national security, and intensifying competition with China.

Implementation Guidance for Federal Agencies

The order directs the director of the Office of Management and Budget, along with key administrators, to formulate guidance for other agencies regarding compliance with the new regulations.

Trump's Stance on Partisan Bias in AI

“Once and for all, we are getting rid of woke,” Trump declared during an AI event. He announced his intention to ban the federal government from procuring AI technology that exhibits partisan bias or ideological agendas, specifically mentioning critical race theory as an example. He asserted that the U.S. government will only engage with AI that prioritizes truth, fairness, and strict impartiality.

The Challenge of Defining Impartiality

Determining what constitutes impartiality or objectivity presents a significant challenge to the effective implementation of the order.

Language is Inherently Not Neutral

Philip Seargeant, a senior lecturer in applied linguistics at the Open University, explained to TechCrunch that true objectivity is unattainable. “One of the fundamental tenets of sociolinguistics is that language is never neutral,” Seargeant stated. “So the idea that you can ever get pure objectivity is a fantasy.”

Trump Administration's Ideological Alignment

Furthermore, the Trump administration’s ideology does not necessarily reflect the diverse beliefs and values held by all Americans. Trump has consistently advocated for reducing funding for initiatives related to climate change, education, public broadcasting, research, social services, community support, agricultural programs, and gender-affirming care, often characterizing these as examples of “woke” or politically motivated government spending.

"Woke" as a Pejorative Term

Rumman Chowdhury, a data scientist and CEO of Humane Intelligence, observed that “Anything [the Trump administration doesn’t] like is immediately tossed into this pejorative pile of woke.”

Vague Definitions in the Executive Order

The definitions of “truth-seeking” and “ideological neutrality” within the published order are somewhat ambiguous, while also being specific in certain areas. “Truth-seeking” is defined as LLMs that “prioritize historical accuracy, scientific inquiry, and objectivity,” while “ideological neutrality” is defined as LLMs that function as “neutral, nonpartisan tools that do not manipulate responses in favor of ideological dogmas such as DEI.”

Potential for Pressure and Interpretation

These definitions allow for broad interpretation and could potentially create pressure on AI companies to align their outputs with the administration’s political agenda. AI companies have generally favored fewer restrictions on their operational practices.

Impact on Department of Defense Contracts

Last week, OpenAI, Anthropic, Google, and xAI secured contracts with the Department of Defense, potentially receiving up to $200 million each to develop agentic AI workflows addressing critical national security concerns.

Compliance and Potential Beneficiaries

It remains unclear which of these companies is best positioned to benefit from the “woke AI” ban, or whether they will fully comply with the new regulations.

Reaching Out to AI Companies

TechCrunch has contacted each of the aforementioned companies for comment and will update this article upon receiving a response.

xAI and Grok's Alignment with the Order

Despite exhibiting its own biases, xAI may be the most aligned with the order, at least initially. Elon Musk has positioned Grok, xAI’s chatbot, as an anti-woke, “less biased,” truthseeker. Grok’s system prompts instruct it to avoid deferring to mainstream sources, seek contrarian information, and even reference Musk’s views on controversial topics. Recently, Grok has generated antisemitic comments and praised Hitler on X, among other harmful posts.

Concerns Over Viewpoint Discrimination

Mark Lemley, a law professor at Stanford University, told TechCrunch that the executive order is “clearly intended as viewpoint discrimination, since [the government] just signed a contract with Grok, aka ‘MechaHitler.’”

Grok's Availability to Government Agencies

In addition to its DOD funding, xAI announced that “Grok for Government” has been added to the General Services Administration schedule, making xAI products available for purchase by all government offices and agencies.

Questioning the Ban's Consistency

“The right question is this: would they ban Grok, the AI they just signed a large contract with, because it has been deliberately engineered to give politically charged answers?” Lemley asked. “If not, it is clearly designed to discriminate against a particular viewpoint.”

AI Outputs Reflect Builders and Training Data

As Grok’s system prompts demonstrate, model outputs are a reflection of both the individuals developing the technology and the data used to train the AI. In some instances, excessive caution among developers and AI trained on internet content promoting inclusivity have resulted in skewed outputs. Google’s Gemini chatbot, for example, faced criticism for depicting a black George Washington and racially diverse Nazis – examples cited in Trump’s order.

Fear of Data Rewriting to Align with Political Agendas

Chowdhury expressed her greatest concern that AI companies will actively revise training data to conform to the prevailing political line. She referenced statements made by Musk prior to launching Grok 4, in which he stated that xAI would use the new model to “rewrite the entire corpus of human knowledge, adding missing information and deleting errors. Then retrain on that.”

Implications for Information Access

This could potentially place Musk in a position to determine what is considered true, with significant implications for how information is accessed.

Judgment Calls in Information Control

However, companies have been making decisions about what information is seen and unseen since the inception of the internet.

David Sacks and Concerns About Ideological Control

Conservative David Sacks, appointed as AI czar by Trump, has voiced concerns about “woke AI” on the All-In Podcast, which co-hosted Trump’s AI announcements. Sacks has accused AI creators of embedding left-wing values into their products, framing his arguments as a defense of free speech and a warning against centralized ideological control.

The Impossibility of True Neutrality

Experts emphasize that achieving unbiased or neutral results is impossible, particularly in a world where even facts are often politicized.

The Subjectivity of Objectivity

“If the results that an AI produces say that climate science is correct, is that left wing bias?” Seargeant questioned. “Some people say you need to give both sides of the argument to be objective, even if one side of the argument has no status to it.”

Contact Information for Tips and Confidential Documents

Got a sensitive tip or confidential documents? We’re reporting on the inner workings of the AI industry — from the companies shaping its future to the people impacted by their decisions. Reach out to Rebecca Bellan at rebecca.bellan@techcrunch.com and Maxwell Zeff at maxwell.zeff@techcrunch.com. For secure communication, you can contact us via Signal at @rebeccabellan.491 and @mzeff.88.