AI Governance Institute
← News

Open-Source Runtime Enforcer Exposes the Gap Between Agent Policy and Practice

What happened

The Conduct open-source project is a runtime governance framework for AI agents that intercepts and enforces compliance policy before an LLM call or shell tool invocation executes, distinguishing it from observability-only solutions that record behavior without blocking it. Released by an independent developer under the sseshachala GitHub account, it ships with more than 20 compliance packs pre-mapped to frameworks including the EU AI Act, NIST Artificial Intelligence Risk Management Framework Playbook, ISO/IEC 42001:2023, HIPAA, PCI DSS 4.0, and SOC 2. Policy integrity is verified cryptographically using signed configuration, and the audit trail uses SHA-256 hash chaining to make tampering detectable. The project is directly relevant to teams managing Model Context Protocol tool integrations, an area where 91.8% of audited MCP servers lack OAuth and documented vulnerabilities have included arbitrary file writes and command execution.

Why it matters

  • ·The tool forces a clarifying question that regulators are increasingly asking: does your agent governance posture enforce policy at runtime, or does it only observe and report? The OWASP Top 10 for Large Language Model Applications and active EU AI Act enforcement both treat preventive controls as a higher standard than logging alone, meaning observability-only programs may face documentation deficiencies during audits.
  • ·Adopting open-source governance tooling into an enterprise AI pipeline creates its own supply chain and integrity risk. The same cryptographic integrity controls Conduct provides for agent policy must be applied to the tool itself, including version pinning, dependency review, and intake assessment under any open-source model or tool intake policy already in place.
  • ·The multi-framework compliance pack approach reflects a market signal about where enterprise expectations are heading: compliance teams are expected to demonstrate simultaneous conformance across several frameworks, and tools that pre-map controls reduce the manual effort of that mapping but also create a single point of failure if the mapping is incomplete or misconfigured.

Stay ahead of stories like this

Get every Global AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Governance controls affected

What to do now

  • ☐Assess whether your current agent governance tooling enforces policy before tool calls execute or only logs behavior after the fact, and document the gap if it exists.
  • ☐Before evaluating or adopting Conduct or similar open-source governance tooling, run it through your open-source intake policy, including dependency review, maintainer vetting, and version-pinning requirements.
  • ☐Map the tool's 20-plus compliance packs against your own multi-framework obligations to identify any gaps or mismatches before treating pre-built mappings as authoritative.
  • ☐Verify that your MCP server and LLM tool-call governance controls include a fail-closed default, and update your agent permission boundary documentation to reflect that standard.
  • ☐Confirm that audit log integrity controls for agent activity include tamper-detection mechanisms such as hash chaining, and test retrieval procedures against your audit trail SLA.

What to watch next

As agent governance is becoming binding across multiple jurisdictions, regulators and auditors are likely to begin distinguishing explicitly between preventive runtime controls and post-hoc logging in their examination guidance. Compliance teams should monitor upcoming enforcement guidance from the EU AI Office and any sector-specific updates from financial and health regulators on what constitutes a sufficient agentic AI control posture. The open-source tooling market for agent governance is developing quickly, and teams should expect the NIST Artificial Intelligence Risk Management Framework Playbook and related standards bodies to address runtime enforcement expectations more specifically in forthcoming updates.

Related Coverage

Research2026-11-08

Coding Agents Expand the Security Boundary Beyond What Traditional Tools Cover

A Forrester analyst commentary published November 5, 2026 argues that AI coding agents introduce a security perimeter that traditional tools cannot assess. The expanded boundary includes AI models, MCP servers, skills, hooks, extensions, and plugins. Compliance teams are advised to maintain inventories of agentic dependencies, require workflow-specific vendor testing, validate AI-generated code, and strengthen access and monitoring controls.

Research2026-11-17

87.5% of Known MCP Vulnerabilities Are Critical or High Severity

Researchers at IEEE SPW 2026 published a curated dataset of 24 validated MCP vulnerabilities drawn from CVE disclosures through February 2026. They found that 87.5% are Critical or High severity, and that two attack types — injection and path traversal — account for 66.7% of cases. The paper also releases a maintained dataset and reproducibility tooling for ongoing MCP security measurement.

Corporate Policy2026-11-15

Google Cloud's Policy MCP Server Lets AI Agents Rewrite Cloud Security Rules

Google Cloud has made its Organization Policy Model Context Protocol (MCP) server generally available. The server allows autonomous AI agents and automated pipelines to inspect, evaluate, and create policies and custom constraints across a Google Cloud environment. Enterprises connecting agents to this server must carefully limit what those agents are permitted to change, and log every evaluation and modification request.