US Executive Order on Safe, Secure, and Trustworthy AI

Summary

On October 30, 2023, President Biden signed Executive Order 14110, marking the most comprehensive federal action on artificial intelligence in US history. This sweeping directive doesn't just offer guidance—it mandates concrete actions from federal agencies, establishes new reporting requirements for AI companies, and sets the stage for a new era of AI oversight. The order leverages the Defense Production Act to require AI companies developing powerful models to share safety test results with the government, while simultaneously directing agencies to create new standards for AI safety, security, and trustworthiness across sectors from healthcare to criminal justice.

The Eight Pillars of Federal AI Action

The Executive Order organizes its requirements around eight strategic priorities:

New Standards for AI Safety and Security - Directs NIST to establish guidelines for red-team testing and requires companies training models with 10^26 flops or more to report results to the Department of Commerce.

Protecting Americans' Privacy - Tasks agencies with developing guidance on privacy-preserving techniques and establishing guidelines for federal agencies' use of commercially available information.

Advancing Equity and Civil Rights - Requires agencies to combat algorithmic discrimination and provides guidance for landlords, federal contractors, and federal agencies to ensure AI doesn't perpetuate bias.

Standing Up for Consumers, Patients, and Students - Creates sector-specific safety frameworks for healthcare AI, establishes principles for AI in education, and directs consumer protection measures.

Supporting Workers - Develops principles for AI's impact on the workforce and creates resources for workers affected by AI-driven changes.

Promoting Innovation and Competition - Streamlines visa processes for AI talent, supports AI research through the National AI Research Resource, and addresses intellectual property questions around AI-generated content.

Advancing American Leadership Abroad - Establishes guidelines for federal agencies' overseas AI work and promotes US AI standards internationally.

Ensuring Responsible Government Use - Requires agencies to inventory their AI use cases, establish Chief AI Officers, and implement minimum practices for AI in government.

Timeline: When Things Actually Happen

Unlike many policy documents, this Executive Order includes specific deadlines that create urgency:

• 90 days: NIST must establish guidelines for conducting safety and security tests
• 120 days: OMB must issue guidance on government AI use and procurement
• 150 days: Department of Labor must produce report on AI's labor market impacts
• 180 days: Multiple agencies must deliver sector-specific guidance (healthcare, education, housing)
• 240 days: Commerce Department must submit report on dual-use foundation models
• 365 days: Several comprehensive frameworks and standards must be completed

These aren't aspirational goals—they're binding directives to federal agencies with real accountability mechanisms.

Who This Resource Is For

AI Companies and Developers developing large-scale models need to understand new reporting requirements, especially the 10^26 flops threshold that triggers mandatory government disclosure of safety testing results.

Federal Contractors across all sectors should prepare for new AI-related requirements in government contracts and potential audits of AI systems used in federal work.

Regulated Industry Players in healthcare, finance, transportation, and other sectors need to track forthcoming agency guidance that will interpret how existing regulations apply to AI systems.

State and Local Government Officials can use this as a roadmap for developing their own AI governance approaches, as many provisions include guidance for non-federal government entities.

Civil Rights and Privacy Advocates will find detailed commitments to addressing algorithmic bias and privacy protection that create new avenues for accountability.

International Partners and Competitors should understand this as the US positioning statement on AI governance and a preview of American approaches to international AI standards.

What Makes This Executive Order Different

This isn't typical government guidance—it's binding federal action with teeth. The use of the Defense Production Act gives the federal government unprecedented authority to compel private companies to share information about their AI development processes. The order also creates new institutional structures, including mandatory Chief AI Officers across federal agencies and new inter-agency coordination mechanisms.

The scope is remarkably broad, touching everything from visa policy for AI researchers to federal procurement rules to international standard-setting. Rather than creating entirely new regulatory frameworks, it largely directs existing agencies to apply their current authorities to AI systems—a pragmatic approach that can move faster than new legislation.

Most significantly, it establishes the federal government as an active participant in AI development rather than just a regulator, through initiatives like the National AI Research Resource and expanded federal AI research programs.

Common Misconceptions

"This only affects big tech companies" - The order's impacts extend to any organization that contracts with the federal government, receives federal funding, or operates in regulated sectors.

"These are just guidelines" - While some provisions create voluntary frameworks, many establish binding requirements for federal agencies and mandatory reporting for AI companies meeting specific thresholds.

"It will slow down AI innovation" - The order includes substantial provisions designed to accelerate American AI leadership, including immigration reforms, research funding, and international cooperation initiatives.

"This replaces the need for Congress to act" - The Executive Order explicitly acknowledges the need for congressional action on AI and is designed as a foundation for, not replacement of, potential AI legislation.