💡 AI Supply Chain Security with AIBOM/ML-BOM for Sam

[github-actions[bot]]

Summary

Extend Sam's security plan workflow with AI-specific supply chain security guidance covering AI Bill of Materials (AIBOM/ML-BOM) generation, model provenance verification, dataset lineage tracking, weight integrity validation, and AI-specific dependency scanning. As organizations move from static SBOMs to active governance and regulatory bodies prepare AI-specific supply chain requirements, teams need structured guidance for securing the AI components of their stack alongside traditional software dependencies.

Market Signal

Cloudsmith's 2026 guide documents the evolution "From Static SBOMs to Agentic Governance" — SBOMs generated but never acted upon are compliance snapshots filed in a digital drawer. SBOM catalogs software dependencies, while ML-BOM extends to AI-specific components: training datasets, base models, fine-tuning data, evaluation benchmarks. AI supply chain attacks include pickle RCE in model weights, embedded backdoors, dependency confusion in transformer packages, and dataset poisoning (AppScale, Pharos Production reports). Procurement and compliance teams are beginning to ask for AI inventories the same way they ask for SBOMs.

User Signal

Sam's existing security plan workflow covers supply chain security (SBOM generation in SPDX/CycloneDX, dependency scanning, container security, SLSA provenance) but exclusively for software dependencies. As BMAD is an AI-native development framework, users building with BMAD are highly likely to incorporate AI/ML components. No existing issues or discussions address AI supply chain security specifically, making this a forward-looking but well-grounded extension of Sam's existing expertise.

Technical Opportunity

Sam's security plan workflow already has a supply chain security section referencing SBOM, SLSA, cosign/Notary. Extending this to cover AIBOM is a natural content expansion. The step covering supply chain security can add conditional sections for AI component inventory, model provenance verification workflows, and dataset lineage tracking. The security plan template can include an "AI Supply Chain" appendix alongside the existing SBOM requirements.

Assessment

Dimension	Score	Rationale
Feasibility	high	Natural content extension to Sam's existing supply chain security section — same principles, new artifact types
Impact	high	AI supply chain attacks are a growing threat vector; regulatory requirements for AI inventories are emerging
Urgency	med	AIBOM standards are maturing but not yet mandated — early guidance positions teams ahead of requirements

Adversarial Review

Strongest objection: AIBOM standards are still maturing — CycloneDX ML-BOM and SPDX AI Profile are not yet universally adopted. Content may be premature and require frequent updates.

Rebuttal: Sam's existing approach to compliance frameworks (SOC2, HIPAA, PCI-DSS, GDPR) already navigates evolving standards by grounding guidance in principles (know your dependencies, verify provenance, track lineage) while referencing specific standards as current best practice. The same approach works for AIBOM: the principle of "know what AI components you're running and where they came from" is stable even as the specific format standards evolve. Early guidance positions teams ahead of regulatory mandates rather than scrambling to catch up.

Suggested Next Step

Add an "AI Supply Chain Security" conditional section to the security plan workflow, covering AIBOM/ML-BOM generation, model provenance verification, dataset lineage, and AI-specific vulnerability scanning. Include a mapping to emerging standards (CycloneDX ML-BOM, SPDX AI Profile) and common AI supply chain threat patterns.

[github-actions[bot]]

Weekly Update

What Changed

• EU AI Act enforcement begins August 2, 2026 — less than two weeks away. Article 11 and Annex IV require technical documentation including AI supply chain transparency, making AIBOM/ML-BOM generation not just a best practice but a regulatory obligation for high-risk AI systems deployed in EU markets.
• CISA and G7 partners released joint SBOM-for-AI guidance — formalizing expectations for AI supply chain transparency across international regulatory bodies. This accelerates enterprise adoption pressure.
• Industry has moved from "visibility era" (static SBOMs) to "governance era" (agentic, continuously updated) — per Cloudsmith's 2026 supply chain security guide. This validates the AIBOM approach of generating bills of materials at build time and maintaining them continuously.
• CycloneDX ML-BOM and SPDX 3.0 remain the two competing specs — Sam's AIBOM workflow should support both formats to maximize compatibility.

Sources: EU AI Act enforcement timeline, CISA/G7 SBOM-for-AI guidance (industrialcyber.co), Cloudsmith 2026 supply chain security guide, Dark Reading SBOM analysis (2026).

Updated Assessment

Dimension	Previous	Current	Delta
Feasibility	med	med	→
Impact	high	high	→
Urgency	med	high	↑ EU AI Act enforcement date is now imminent (August 2, 2026). AIBOM generation moves from "best practice" to "regulatory requirement" for EU market access.

Recommendation

Advance urgently — The EU AI Act enforcement deadline creates a concrete business case. This idea should be prioritized alongside the new EU AI Act Compliance Readiness Mapping proposal (#413) as complementary capabilities: #413 maps compliance requirements broadly, while this AIBOM workflow produces specific supply chain artifacts required by Annex IV.