AI BOMs Are the Missing Link in Enterprise Supply Chain Transparency
What happened
In a podcast interview published by Software Transparency Is Changing: Allan Friedman on SBOM, AI BOMs and What Comes Next, cybersecurity policy expert Allan Friedman explained why a Software Bill of Materials, or SBOM (a structured list of every software component an application depends on), is only useful when treated as a live input. It must feed vulnerability management and procurement decisions, not sit as a document filed and forgotten. Friedman highlighted the Vulnerability Exploitability eXchange, or VEX, as a companion standard. VEX tells organizations whether a known flaw in a listed component is actually reachable in their specific deployment. This helps teams prioritize which issues genuinely need attention. He then described how AI systems break the traditional SBOM model. An AI application may depend on a third-party model, training data, and external data fed in at runtime. It may also rely on surrounding infrastructure. None of these appear in a standard software component list. The hardest problem, Friedman argued, is agentic AI. A model may create a tool, execute an action, and leave no persistent record of what it did or accessed. The solution he outlined is not a single document format. It is a convergence of SBOM concepts, detailed activity logs, and third-party risk management practices sufficient to reconstruct what happened after an incident.
Why it matters
- ·Organizations that cannot inventory AI model origins, training data sources, and runtime data inputs will struggle to assess exposure when a supplier is compromised. This mirrors the Log4j problem, but standard software inventories miss the AI-specific components entirely.
- ·Regulators and frameworks demanding AI transparency, including the EU AI Act (Regulation (EU) 2024/1689), increasingly expect organizations to document what is inside their AI systems. An SBOM that stops at software libraries and omits model provenance or external data feeds will not satisfy that expectation.
- ·Agentic AI creates an audit trail gap that static inventory tools cannot close. When an AI agent creates a temporary tool, takes an action, and terminates, traditional logging and change-control records may contain no trace of what occurred. This makes forensic investigation after an incident much harder.
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.
Recent issues
- State-linked attackers and criminal operators are actively weaponizing AI tools as entry points into enterprise systems, turning internal platforms and open-source frameworks into live attack vectors.5 Nov
- Three separate incidents this week each created a new compliance action item: verify your AI vendors' data sourcing, demand agent monitoring coverage guarantees, and audit trusted app identities in your cloud environment.29 Oct
Free every Thursday. Unsubscribe anytime.
Governance controls affected
What to do now
- ☐Ask your engineering and procurement teams whether AI vendors can produce a bill of materials that covers not just software libraries but also the AI model itself, its training data origins, and any external data sources the model queries at runtime.
- ☐Review your existing software inventory and vulnerability management processes to confirm they can ingest AI component data, including model version, model source, and runtime data dependencies, not just traditional software package names.
- ☐Require vendors who supply AI-enabled products to provide VEX statements alongside any vulnerability disclosures, so your team can determine whether a reported flaw is actually reachable in your specific deployment rather than treating every alert as equally urgent.
- ☐Identify any AI agents in production that create temporary tools or take autonomous actions, and confirm that your activity logging captures enough detail to reconstruct what each agent did, what data it accessed, and what external systems it contacted.
- ☐Update your third-party AI vendor due diligence questionnaire to ask specifically about AI BOM capability: can the vendor tell you what model version is running, where it originated, and what data sources it draws on at runtime?
What to watch next
Regulatory requirements for AI component transparency are likely to sharpen as the EU AI Act (Regulation (EU) 2024/1689) moves from broad principles to enforcement practice. Bodies such as CISA and NIST continue expanding SBOM guidance to cover AI-specific components. Industry standards for AI BOMs are still forming, so compliance teams should monitor NTIA and NIST working groups for AI-specific SBOM schema guidance that could become a de facto requirement in procurement contracts. The emergence of agentic AI as an audit trail problem will also draw increasing regulatory attention, particularly as incident investigations begin revealing gaps in what enterprises can actually reconstruct after an AI-driven event.