AI Compliance and Software Supply Chain Security: The Trace-AI Metadata Model

Updated: Oct 9

The Compliance Blind Spot in the AI Era

Artificial intelligence has revolutionized how we build software, but the tools we use to prove it’s secure are stuck in the past. Today’s applications are complex assemblies of pre-trained models, microservices, and countless open-source dependencies. While this accelerates innovation, it also creates a critical challenge for software supply chain security: how do you secure what you can't see?

Traditional vulnerability scanners were not built for this new reality. They require access to source code and assume a static environment, an assumption that collapses with opaque AI models and cloud-native components that change daily. This leaves a dangerous blind spot in your security and compliance posture.

At Zerberus, we believe the answer isn't to look deeper into the code, but to understand the signals around it. That’s why we built Trace-AI, a platform that uses metadata analysis to deliver continuous, verifiable evidence for AI compliance.

Why Metadata Is the New Frontier for Software Supply Chain Security

Think of metadata as the digital fingerprint of your software components. While source code reveals what a component does, metadata reveals its history, behaviour, and trustworthiness. It tells you who published a package, its update cadence, its maintainer's reputation, and where it originates.

Code scanners hunt for known CVEs - a reactive approach. In contrast, metadata analysis proactively detects behavioural anomalies that often precede a compromise. For example:

Suspicious Maintainer Changes: A package that suddenly changes ownership after years of stability.
Typosquatting Attacks: A new library whose name subtly mimics a popular, trusted one.
Dependency Abandonment: A critical dependency that is no longer maintained but remains in your production environment.

These events leave clear traces in metadata long before a vulnerability is exploited. By monitoring these signals, Trace-AI provides predictive insights, turning your software supply chain security from a reactive chore into a proactive strategy.

From Academic Research to a Practical SBOM

The core of Trace-AI is built on pioneering research from Royal Holloway, University of London. There, the Zerberus team developed the ZSBOM, a metadata-only Software Bill of Materials (SBOM).

The research confirmed a powerful hypothesis: we could assess software risk more effectively using behavioural patterns found in metadata than by just scanning code for known flaws. This academic framework became the blueprint for Trace-AI’s engine, designed to automate evidence generation for major compliance frameworks like ISO 27001, SOC 2, and the upcoming EU Cyber Resilience Act (CRA).

How Trace-AI Delivers Continuous Compliance via Metadata Analysis

Trace-AI establishes trust without ever needing to access your proprietary source code or AI model weights. It operates across four logical layers to create a living map of your software ecosystem:

Source Graph: First, it maps every dependency, AI model, container image, and API in your environment, creating a comprehensive inventory.
Context Engine: Each component is then enriched with contextual metadata, including its origin, maintainer trust score, repository activity, and version history.
Inference Layer: The platform’s AI model analyses these metadata signals to identify and predict risks like typosquatting, dependency abandonment, or version drift.
Copyleft License Violations: A package with a GPL-3.0 or AGPL-3.0 license that could force your proprietary code disclosure.
Policy Engine: Finally, these insights are translated directly into auditable compliance artefacts, automating the evidence required for AI compliance and regulatory reporting.

A traditional scanner is like an X-ray, looking for a known fracture. Trace-AI is like a continuous EKG, monitoring the heartbeat of your supply chain to detect irregularities before they become a critical issue.

Automating Evidence for the Cyber Resilience Act, ISO 27001, and SOC 2

This metadata-driven approach fundamentally transforms compliance from a periodic, manual effort into an automated, ongoing process. For builders and auditors, Trace-AI delivers:

Continuous Assurance: Maintain compliance without halting development or release cycles.
Automated Evidence Generation: Automatically map security controls to requirements for the Cyber Resilience Act, NIS2, ISO 27001, and SOC 2.
Privacy-Preserving Security: Prove the integrity of your software and AI models without exposing intellectual property.
Predictive Visibility: Gain a complete view of risk across your entire AI and software supply chain.

Aspect	Traditional Scanners	Trace-AI Metadata Model
Code Access	Required	Not Mandatory
Detection Method	CVE-based (Reactive)	Behavioural & Exploit-Aware (Proactive)
Primary Focus	Vulnerabilities in Code	Trust & Integrity of Supply Chain
Compliance Output	Manual Reports	Automated, Continuous Attestations
Privacy	Intrusive	Preserved by Design

A Simple Example: Detecting a Compromised Package

Imagine a Python package in your stack suddenly changes maintainers and pushes three updates in a week after years of inactivity. A conventional scanner sees no CVEs and reports nothing.

Trace-AI, however, instantly flags this behavioural anomaly through its metadata analysis. It triggers an alert, re-evaluates the component's trust score, and generates a signed evidence pack for your compliance dashboard - all before a malicious payload can be delivered. This is the power of proactive software supply chain security.

The Road Ahead: Launching Trace-AI

This metadata-driven model is the core of Trace-AI, the next product from Zerberus. We are operationalising the ZSBOM framework to embed continuous compliance directly into your CI/CD pipelines.

Early adopters are already using Trace-AI to monitor dependencies, establish provenance, and ensure AI model integrity in real-time. We are preparing for our public launch on Product Hunt and are now opening early-access invitations.

Be among the first to experience how metadata analysis can transform your AI compliance and security posture. Signup at https://zerberus.ai (or) https://app.zerberus.ai

The Future is Verifiable

The future of compliance will be written in metadata, not audit reports. When every component in your pipeline can prove its own integrity, security becomes a natural outcome of how you build. Trace-AI is designed to make that future a reality.