US Executive Order directs agencies to avoid ideologically biased AI models
A new Executive Order signed on July 23, 2025, directs U.S. federal agencies to procure only those large language models (LLMs) that meet newly defined standards of truthfulness and ideological neutrality, citing concerns over the influence of diversity, equity, and inclusion (DEI) principles on AI outputs.
On 23 July 2025, President Donald J. Trump signed an Executive Order establishing new requirements for the use of artificial intelligence (AI) within the federal government. The order mandates that agencies may only procure large language models (LLMs) that meet two core ‘Unbiased AI Principles’: truth-seeking and ideological neutrality.
These principles require AI systems to prioritise factual accuracy, scientific rigour, and objectivity, while explicitly avoiding the incorporation of what the order describes as ‘ideological dogmas,’ such as diversity, equity, and inclusion (DEI). It criticises practices such as altering historical figures’ characteristics or prioritising certain narratives in AI-generated responses.
The Office of Management and Budget (OMB) is tasked with issuing implementation guidance within 120 days. This will influence not only new AI contracts but, where feasible, existing ones as well. Exceptions are allowed for national security systems, and the order provides flexibility for vendors in how they comply, aiming to balance transparency with protection of proprietary data.
Why this matters for civil society
Although the order applies to federal procurement, its implications reach beyond government use. Public sector standards often influence broader AI development norms. Some experts in the field have already expressed concerns that framing DEI as inherently biassed may restrict access to information and limit the ability of AI tools to reflect diverse perspectives.
By narrowing the acceptable ideological range in AI outputs, the order may affect how key social issues, such as race, gender, and inequality, are represented or discussed in publicly funded systems. This raises concerns about digital inclusivity, freedom of expression, and the right to access balanced information. Civil society groups play a critical role in monitoring these developments to ensure that governmental technology use remains transparent, accountable, and aligned with democratic values.
This Executive Order builds on the 2020 directive (EO 13960) promoting trustworthy AI in federal operations. However, its specific framing and constraints signal a shift in how the US government defines objectivity and neutrality in AI.
