A case for how to shape ‘ingredient lists’ for AI models

A case for how to shape ‘ingredient lists’ for AI models
A policy paper from the Institute for Security and Technology argues that AI software bills of materials (AIBOMs) could reduce cyber risk and improve transparency by creating a shared inventory of models and datasets used in AI systems. The authors say AIBOM policy needs both supply-side data collection and demand-side requirements, such as industry mandates, government regulations, or contracting conditions, before broad adoption can work. #AIBOM #InstituteForSecurityandTechnology #AllanFriedman #NickLeiserson

Keypoints

  • The paper promotes AIBOMs as a way to improve AI supply chain transparency.
  • It says foundational work is needed before AIBOMs can be widely implemented.
  • AIBOMs should track models, datasets, and other AI development inputs.
  • The policy approach must address both data supply and data demand.
  • The authors cite possible mandates, regulations, or contracting rules to drive adoption.

Read More: https://cyberscoop.com/ai-bill-of-materials-policy-roadmap/