Response to OSTP RFI on the Development of a 2025 National Artificial Intelligence Research and Development Strategic Plan

This post contains IAPS’s comments on the Office of Science and Technology Policy’s Request for Information on the Development of a 2025 National Artificial Intelligence (AI) Research and Development (R&D) Strategic Plan. These comments focus on ways the AI R&D Strategic Plan can advance responsible AI innovation while maintaining America’s technological leadership.

Executive Summary

Artificial intelligence represents the most significant technological opportunity of our time, with potential to drive trillions in economic growth1 and revolutionize scientific discovery.

However, this opportunity comes with substantial challenges. Foreign adversaries may leverage AI's capabilities to threaten U.S. national security and economic competitiveness. These adversaries will also likely target our AI research, supply chains, and infrastructure.

Beyond adversarial threats, as AI systems become more autonomous, they may become deceptive or even misaligned, creating additional risks.4 The 2025 National AI R&D Strategic Plan offers a chance to guide federal research investments crucial for navigating these challenges and securing AI's benefits.

IAPS's work makes it uniquely positioned to recommend goals for the 2025 plan:

• Research and Reports: IAPS has published detailed reports covering infrastructure and compute, AI agents, and cybersecurity.

• R&D Survey: IAPS recently surveyed experts across 105 AI reliability and security research areas to identify promising research prospects for strategic AI R&D investment.

Here we outline four research goals that the 2025 National AI R&D Strategic Plan should pursue:

I. Strengthen AI Model Security, Reliability, and Capability Assessments

II. Advance the Security of AI Supply Chains and Infrastructure

III. Develop Methods to Monitor Adversary AI Capabilities

IV. Evaluate, Secure, and Support Multi-Agent Systems