The US Executive Order on AI and What Happened Next

Executive Order 14110, Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, ran over 100 pages. It set requirements across eight agencies and touched almost every part of the federal government's relationship with AI.

Key provisions

• Reporting requirements for training runs above 10^26 FLOPs (10^23 for biological models)
• Safety test results sharing with the federal government for covered models
• NIST tasked with developing AI safety standards and the AI Safety Institute
• Export controls and KYC rules for US IaaS providers serving foreign entities
• Civil rights, labor, and housing AI guidance
• Agency AI use policies and Chief AI Officer roles

What changed in January 2025

On January 20, 2025, President Trump rescinded EO 14110 on his first day in office. In its place came Executive Order 14179, Removing Barriers to American Leadership in Artificial Intelligence (January 23, 2025), which emphasizes deregulation and US competitiveness. A new AI Action Plan followed in mid-2025.

1. 1Biden EO 14110: safety-first framing, reporting requirements, civil rights focus
2. 2Trump EO 14179: competitiveness-first framing, deregulatory, open to state preemption
3. 3NIST AI Safety Institute survived under new name/mandate: US Center for AI Standards and Innovation (CAISI)
4. 4Reporting requirements were substantially loosened but some frameworks persisted

“An executive order is a statement of priorities. A law is a constraint. We have had many of the first and few of the second.”

The big idea: US AI policy oscillates with each administration. The real baseline is what agencies build while they can — NIST standards, procurement rules, voluntary commitments — because those persist through political change.