NSF Awards $587K to Redesign Web Infrastructure for AI Agents

NSF Awards $587K to Redesign Web Infrastructure for AI Agents

The National Science Foundation has awarded a five-year, $587,043 CAREER Award to Washington University in St. Louis researcher Umar Iqbal to develop new web infrastructure designed for AI agents that can autonomously take actions online.

Iqbal, an assistant professor of computer science and engineering in the McKelvey School of Engineering, will focus on creating frameworks that address security, privacy, and reliability challenges as AI systems increasingly interact with websites on behalf of users.

Why it matters

The web was built decades ago to retrieve and display text documents for human readers. As AI agents gain the ability to browse sites, fill forms, make purchases, and execute tasks without human supervision, the underlying architecture lacks safeguards for these autonomous actions. Without new infrastructure, organizations face risks ranging from unauthorized transactions to data breaches as AI systems operate in environments never designed for machine autonomy.

The Infrastructure Gap

According to Iqbal, the current web architecture cannot adequately handle the shift from passive document retrieval to active, autonomous operations. The original internet design assumptions—that humans would directly control all actions—no longer hold as generative AI systems create content and agentic systems execute tasks independently.

The research will address fundamental questions about how websites can verify AI agent identities, enforce permission boundaries, and maintain audit trails when machines rather than humans initiate actions. These technical challenges have business implications for any organization deploying or interacting with AI agents online.

Research Focus Areas

Iqbal brings expertise in machine learning, artificial intelligence, and computer security and privacy to the project. The CAREER Award, one of NSF's most prestigious grants for early-career faculty, will support development of protocols and standards that allow AI agents to reliably use web services while maintaining security controls.

The five-year timeline reflects the scope of redesigning core internet infrastructure to accommodate a fundamentally different model of web interaction. Rather than patching existing systems, the research aims to establish new architectural principles suited for an era of autonomous AI operations.

Broader Implications

As enterprises increasingly deploy AI agents for customer service, data gathering, and business process automation, the lack of appropriate infrastructure creates operational risks. The research could inform industry standards and best practices for organizations building or hosting AI-powered services.

The work also has implications for regulatory frameworks, as policymakers grapple with accountability questions when AI systems act independently online. Clear technical boundaries and verification mechanisms could help establish clearer lines of responsibility.

This story was first reported by Washington University in St. Louis.