An SBoM for an AI application catalogs every component: base models, fine-tuned checkpoints, libraries (e.g., PyTorch, Transformers), training datasets, and inference dependencies. This digital provenance creates transparency, allowing you to trace the origin of every part of your system. Without an SBoM, you cannot effectively manage vulnerabilities, comply with regulations like the EU AI Act, or audit your AI supply chain for risks like poisoned data or compromised models.
Guide
Setting Up a Software Bill of Materials (SBoM) for AI Applications
A practical guide to generating a comprehensive SBoM for AI applications. Learn to inventory models, datasets, and dependencies using Syft, format with SPDX/CycloneDX, and automate security scans with Grype in your CI/CD pipeline.
A Software Bill of Materials (SBoM) is a formal, machine-readable inventory of all components in a software application. For AI systems, this is critical for security, compliance, and operational integrity.
You generate an SBoM by integrating automated tools like Syft and Grype into your CI/CD pipeline. These tools scan container images and directories to identify components, which are then formatted into standard schemas like SPDX or CycloneDX. The final SBoM is a JSON or XML file that can be shared with partners, fed into GRC workflows, or used to trigger security scans when new vulnerabilities are disclosed, forming the foundation for a provenance-aware AI development platform.
SBoM FUNDAMENTALS
Key Concepts: AI-Specific SBoM Components
A Software Bill of Materials for an AI application must catalog unique components beyond traditional software, including models, data, and specialized dependencies. This is the foundation for supply chain security and compliance.
02
Training & Evaluation Datasets
The provenance of data is a critical SBoM entry. AI models are defined by their training data, making its lineage a security and compliance requirement.
- Document: Dataset name, version, source (e.g., web crawl, proprietary), licensing, and a cryptographic hash of the dataset snapshot.
- Include: Any data preprocessing or augmentation scripts as dependencies.
- Why: This enables auditing for copyright, bias, or data poisoning issues. Learn more about data provenance in our guide on Establishing a Provenance Verification Framework for Training Data.
EXPLORE
05
SBoM Format & Standards
The structured schema used to encode the component list. Using a standard ensures interoperability with security tools and partners.
- SPDX (ISO/IEC 5962): The most comprehensive standard, ideal for capturing licenses and complex relationships between components.
- CycloneDX: Often favored in DevOps for its smaller footprint and native support for vulnerability reporting.
- Choose: Use SPDX for rigorous compliance needs. Use CycloneDX for integration with CI/CD and existing software supply chain tools. Our guide on Building an SBoM Generation Pipeline covers implementation.
PREREQUISITES
Step 1: Install SBoM Tooling (Syft & Grype)
This step installs the core tools for generating and analyzing a Software Bill of Materials (SBoM), establishing the foundation for AI supply chain security.
A Software Bill of Materials (SBoM) is a formal, machine-readable inventory of all components in your application. For AI systems, this includes base models, fine-tuned checkpoints, Python libraries, and system dependencies. We use Syft to generate the SBoM and Grype to scan it for known vulnerabilities. This duo automates the creation of a provenance-aware component list, which is critical for compliance and security audits under frameworks like the EU AI Act.
Installation is straightforward via package managers. For Linux/macOS, use the install scripts. For containerized environments, pull the Docker images. Verify the install by running syft version and grype version. These tools will form the core of your automated SBoM pipeline, which you will later integrate into CI/CD as part of your broader MLOps and Model Lifecycle Management for Agents.
STANDARD SELECTION
SBoM Format Comparison: SPDX vs. CycloneDX
A feature-by-feature comparison of the two dominant SBoM formats to help you choose the right standard for your AI application's supply chain security.
| Feature / Capability | SPDX | CycloneDX |
|---|---|---|
Primary Standard Body | Linux Foundation | OWASP Foundation |
Core Document Format | JSON, YAML, RDF, Tag-Value | JSON, XML |
AI/ML Component Support | Explicit fields for datasets, models, and training runs | Extended via external 'bom-ref' and custom types |
License Clarity | Dedicated license expression syntax (SPDX License List) | Supports SPDX IDs and named licenses |
Vulnerability Integration | Links to external sources (e.g., CVE) | Can embed VEX (Vulnerability Exploitability eXchange) data |
Cryptographic Signing | Built-in specification for signing the SBoM document | Relies on external standards like JWS or code signing |
Tooling Ecosystem | Focused on software composition analysis (SCA) | Broad, with strong DevOps/CI/CD and cloud-native integration |
Adoption & Governance Fit | Common in open-source and regulatory (e.g., NTIA) contexts | Prevalent in enterprise DevOps and cloud security |
Enabling Efficiency, Speed & Accuracy
Intelligent Analysis, Decision & Execution
We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.
Talk to Us
Search across company data
Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.
Useful when people spend too long searching or get different answers from different systems.
Enterprise searchRAGPermissions
Read more
Automate internal workflows
Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.
Useful when repetitive work moves across multiple tools and teams.
AI agentsWorkflow automationGovernance
Read more
Add AI to products and internal tools
Build assistants, guided actions, or decision support into the software your team or customers already use.
Useful when AI needs to be part of the product, not a separate tool.
AI integrationDecision supportModel routing
Read moreSBoM IMPLEMENTATION
Common Mistakes
Generating a Software Bill of Materials (SBoM) is a foundational step for AI supply chain security, but developers often stumble on the same pitfalls. This guide addresses the most frequent errors and provides clear solutions to ensure your SBoM is accurate, actionable, and integrated.
Standard software SBoM tools often fail to recognize AI-specific artifacts. An SBoM must include all components that constitute your application's "supply chain," which for AI extends beyond libraries to:
- Base and fine-tuned model checkpoints (e.g.,
.safetensors,.binfiles) - Training and validation datasets
- Configuration files (e.g.,
adapter_config.jsonfor LoRA) - Inference server dependencies (e.g., vLLM, TensorRT)
How to fix it: Use AI-aware SBoM generators. Tools like Anchore Syft can be extended with custom catalogers, or you can use spdx-sbom-generator with a pre-scan script to hash model files. The key is to explicitly define and scan the directories containing these non-standard artifacts.
For a complete guide on building this pipeline, see How to Build an SBoM Generation Pipeline for AI Supply Chains.
About the author
Prasad Kumkar
CEO & MD, Inference Systems
Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.
His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.
LinkedIn
Limited slotsGet a Free AI Consultation
Partnered with leading AI, data, and software stack.
How We Work
Custom AI workflows for your Business
One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us