Squares and Rectangles: Why AI Agents Can’t Be Treated Like Regular Workloads

2026-07-16 · Source: Aembit · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, medium

Summary

AI agents, while software, present unique identity and access management challenges that traditional workload models fail to address, according to an article published in July 2026. Unlike deterministic microservices, AI agents exhibit dynamic, non-deterministic behavior, selecting tools and resources at runtime based on task context. This unpredictability creates "scope problems" where credentials are either over-permissioned or too restrictive, "standing access problems" due to static credentials not adapting to evolving agent behavior, and "attribution problems" where logs lack sufficient context for compliance. The article advocates for a "blended identity" approach, which treats AI agents as a distinct identity class, combining agent and user context. This model issues runtime-scoped, task-specific, expiring credentials, evaluated against policies, and ensures comprehensive logging for auditability.

Key takeaway

For MLOps Engineers or AI Architects deploying AI agents, recognize that your existing workload identity models are insufficient. Your agents' dynamic, non-deterministic behavior necessitates a "blended identity" approach, issuing runtime-scoped, task-specific, expiring credentials. Implement this new security framework proactively to avoid critical identity and access gaps, ensuring auditability and policy control without delaying agent rollout. Relying on static service accounts will expose your systems to unpredictable risks.

Key insights

AI agents require a distinct "blended identity" security model, not traditional workload or user identity, due to their dynamic, non-deterministic behavior.

Principles

• AI agents form a distinct identity class.
• Agent access patterns are dynamic and inferred at runtime.
• Static credentials create standing access risks for agents.

Method

Blended identity issues runtime-scoped, task-specific, expiring credentials. Access is policy-evaluated based on agent, resource, and conditions, with all decisions logged for audit and compliance.

In practice

• Issue credentials at runtime, scoped to specific tasks.
• Expire agent credentials once tasks are complete.
• Log all access decisions with agent and user context.

Topics

• AI Agents
• Identity and Access Management
• Workload Identity
• Blended Identity
• Runtime Security
• Access Control

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, MLOps Engineer, AI Architect

Related on AIssential

• Secure Agentic Access: Authentication and Authorization for AI Agent Workloads — Secure agentic access requires unique workload identities, runtime attestation, and task-scoped authorization, moving beyond traditional IAM.
• 10 Identity Security Vendors for AI Agents: Strengths, Tradeoffs and How They Fit — AI agents necessitate a multi-faceted identity security approach, integrating workload IAM, NHI governance, and runtime enforcement beyond traditional human-centric systems.
• No AI Agent Without Identity (Part 2): Building the Layered Identity Model — AI agent identity requires a layered model to balance scalability with robust governance and attribution.
• What is Agentic Security Runtime? Securing AI Agents — Agentic runtime security uses dynamic, session-bound credentials and user identity for secure AI agent interactions.
• Behavioral Credentials: Why Static Authorization Fails Autonomous Agents — Autonomous agents require continuous behavioral attestation, not static authorization, to ensure operational trust and mitigate drift.
• Your agents are using your credentials, and that is the problem — AI agents' non-deterministic behavior breaks traditional static identity and access management models, requiring runtime authority.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Aembit.