Aegis: Towards Governance, Integrity, and Security of AI Voice Agents

2026-02-07

AI Summary

This work addresses critical security threats—such as privacy leakage, privilege escalation, and resource abuse—faced by AI voice agents in high-risk environments, where existing research lacks a systematic evaluation framework. To bridge this gap, the authors propose Aegis, the first comprehensive red-teaming framework tailored for voice agents. Aegis simulates real-world deployment pipelines to construct structured adversarial scenarios, integrating adversarial testing, risk modeling, access control analysis, and behavioral monitoring to holistically assess risks across governance, integrity, and security dimensions. Empirical evaluations in banking, IT support, and logistics settings reveal that even with stringent access controls, voice agents remain vulnerable to behavior-level attacks that circumvent defenses; notably, open-source models exhibit heightened susceptibility, underscoring the urgent need for multi-layered defense mechanisms.

Technology Category

• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness
• Multiagent Systems: Adversarial Agents
• Humans and AI: AI for Accessibility

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Security and Privacy: Security and privacy of machine learning and AI applications
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

With the rapid advancement and adoption of Audio Large Language Models (ALLMs), voice agents are now being deployed in high-stakes domains such as banking, customer service, and IT support. However, their vulnerabilities to adversarial misuse still remain unexplored. While prior work has examined aspects of trustworthiness in ALLMs, such as harmful content generation and hallucination, systematic security evaluations of voice agents are still lacking. To address this gap, we propose Aegis, a red-teaming framework for the governance, integrity, and security of voice agents. Aegis models the realistic deployment pipeline of voice agents and designs structured adversarial scenarios of critical risks, including privacy leakage, privilege escalation, resource abuse, etc. We evaluate the framework through case studies in banking call centers, IT Support, and logistics. Our evaluation shows that while access controls mitigate data-level risks, voice agents remain vulnerable to behavioral attacks that cannot be addressed through access restrictions alone, even under strict access controls. We observe systematic differences across model families, with open-weight models exhibiting higher susceptibility, underscoring the need for layered defenses that combine access control, policy enforcement, and behavioral monitoring to secure next-generation voice agents.

Problem

• voice agents
• adversarial misuse
• security evaluation
• privacy leakage
• privilege escalation

Innovation

• red-teaming framework
• voice agent security
• adversarial scenarios
• behavioral attacks
• Audio Large Language Models