On July 23, 2025, President Donald Trump issued Executive Order 14319, which has the stated purpose of preventing the federal government from procuring A.I. “models that sacrifice truthfulness and accuracy to ideological agendas.” The order specifically targets models that incorporate principles of diversity, equity, and inclusion (DEI), asserting that such frameworks may compromise factual accuracy and reliability.
Under the executive order, federal agencies can procure large language models (LLMs) only if they: (1) are “truthful in responding to user prompts seeking factual information or analysis,” “prioritiz[ing] historical accuracy, scientific inquiry, and objectivity, and [] acknowledg[ing] uncertainty where reliable information is incomplete or contradictory,” and (2) are “neutral, nonpartisan tools that do not manipulate responses in favor of ideological dogmas such as DEI.”
Although the executive order applies only to federal procurement, A.I. companies may be incentivized to apply these requirements to their products more broadly to avoid missing out on sizable federal contracts. Further, the impact of these changes could have unintended effects. As recent high-profile failed updates to some of the most widely used chatbots and image generators demonstrate, changes “under the hood” of these systems can have unintended consequences as they are rolled out.
To the extent this executive order or similar efforts affect the operation of these platforms, it could impact...
Russian propaganda is spreading false statements that the account of the head of the Main Intelligence Directorate of the Ministry of Defense of Ukraine, Kyrylo Budanov, was allegedly created with...