Janus: a Playground for User-Involved Agentic Permission Management

2026-07-01 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

Janus is a publicly available playground system designed to implement and evaluate user-involved permission management in AI agentic systems. Released on 2026-07-01, Janus addresses the underexplored role of users in autonomously executing tool calls. It comprises Janus-Core, a modular agentic system supporting diverse permission management designs, and Janus-Harness, an automated evaluation framework. Grounded in a conceptual model defining key design axes for user involvement, the system implements six permission assistants. Evaluations across three scenarios and three synthetic responders demonstrate that user input significantly strengthens privacy and security, AI augmentation can reduce cognitive load, and system designs must account for realistic user behavior like permission fatigue. The findings indicate no single design performs optimally across all contexts, advocating for a principled, context-sensitive approach.

Key takeaway

For AI Security Engineers designing or deploying agentic systems, you must prioritize robust user-involved permission management. Your designs should integrate AI augmentation to reduce user cognitive load while actively mitigating permission fatigue, as no single approach is universally optimal. Utilize the publicly available Janus playground to implement and evaluate diverse permission assistant designs, ensuring your systems effectively balance security, privacy, and user experience.

Key insights

User involvement is critical for agentic permission management, requiring designs that balance security, cognitive load, and fatigue.

Principles

• User input strengthens agentic system privacy and security.
• AI augmentation can reduce user cognitive load.
• Account for user permission fatigue in system design.

Method

Janus-Core provides a modular agentic system, while Janus-Harness offers an automated evaluation framework for user-involved permission management designs, guided by a conceptual model of design axes.

In practice

• Implement AI augmentation for user permission decisions.
• Design systems to mitigate user permission fatigue.
• Evaluate permission designs across diverse scenarios.

Topics

• Agentic Systems
• Permission Management
• User Involvement
• AI Security
• Privacy
• Permission Fatigue
• Evaluation Frameworks

Best for: Research Scientist, CTO, VP of Engineering/Data, AI Scientist, AI Engineer, AI Security Engineer

Related on AIssential

• Nothing To Detect — Agentic AI's primary security threat is authorized malicious action, shifting focus from detection to governance of permission models.
• Securing our codebase with autonomous agents — Autonomous agents can significantly scale codebase security by automating vulnerability identification and remediation.
• IAM for AI: 4 Steps to Secure and Futureproof Agentic Systems — A four-step maturity model enhances AI agent IAM by addressing accountability, least privilege, abuse prevention, and data safeguarding.
• How Amazon uses agentic AI for vulnerability detection at global scale — Agentic AI systems can dramatically accelerate vulnerability detection rule generation while maintaining high precision through specialized agents.
• Zero Trust Doesn't Fully Solve the Agentic AI Problem — Agentic AI's autonomous nature creates security gaps that Zero Trust alone cannot fully address.
• RSA recap, the LiteLLM breach, and the quest to fix AI agent security — Securing agentic AI requires isolating workflows and dynamic identity management, as traditional IAM fails against non-deterministic threats.

Open in AIssential →

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