The New U S Federal AI Policy Demands Government And Private Sector Tech Leaders Embrace Responsible And Explainable AI

Introduction

Beneath the headlines of geopolitical and geoeconomic volatility, a significant transformation is unfolding in the public sector. The U.S. Federal Government has introduced a new AI policy, marked by Executive Order 14179 and subsequent OMB memoranda (M-25-21 and M-25-22). This policy shifts from government-driven AI innovation to reliance on commercially developed AI, accelerating the "algorithmic privatization" of government services.

The Stakes of Algorithmic Privatization

Historically, privatization involved transferring tasks and personnel from public to private hands. Now, government functions are increasingly delegated to commercially maintained algorithms, large language models, and soon, AI agents and Agentic systems. This raises critical questions:

• Who does a commercially provided AI agent optimize for—the contracting agency, the supplier, or its own evolving model?
• Can AI agents from different suppliers coexist in the same service area? Who governs them?
• What happens when rebidding the AI agent supply relationship? Can knowledge transfer occur, or will it create monopolies?

Risks and Past Failures

The promise of efficiency and innovation comes with substantial risks. Examples of predictive AI failures include:

• Michigan’s MiDAS Unemployment System: A commercially supplied AI wrongly accused thousands of fraud, leading to costly taxpayer-funded compensation.
• Thomson Reuters’ ‘Fraud Detect’ System: This AI erroneously flagged legitimate claims in California, triggering lawsuits and reputational damage.