Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

A Chain Is Only as Strong as Its Weakest Link: A Scoping Review of System Integration Audits in AI

Source

arXiv

arxiv.org/abs/2608.04921v1

Read the source

Source

Plain-Language Brief

A study reviews 58 AI audits that focus on system integration, finding that while few measures address integration-specific risks and significant gaps remain in traditional audit expectations, integration can be categorized across three sites: inter-component, system-environment, and multi-system. The review calls for the AI community to prioritize system integration as a core strategy for addressing AI risk.

What the Document Says

  • AI systems integrated into diverse interfaces and applications require audits beyond model-centric approaches due to risks arising from interactions among components and deployment environments.
  • The role of system integration in AI auditing is underexplored, with scoping review of 58 such audits conducted by analyzing their elements, actors, enablers, and constraints.
  • Three sites of integration were identified: inter-component, system-environment, and multi-system, each serving functions of risk exploration, determination, coordination, and procedural regularity.
  • Integration audits assess qualities specific to system integration, including compatibility, completeness, and oversight.
  • The study finds that while gaps remain in traditional audit expectations, integration can help capture failures across components, environments, and systems beyond the reach of component-level evaluation.
  • The review calls on the AI community to prioritize system integration as a core strategy for addressing AI risk and develop audit practices capable of capturing failures across various levels.

Key Concepts Mentioned

  • AI audits
  • System integration
  • Inter-component
  • System-environment
  • Multi-system
  • Risk exploration
  • Risk determination
  • Coordination
  • Procedural regularity
  • Compatibility
  • Completeness
  • Oversight

Claims Made by the Document

  • Few existing measures target integration-specific risks in AI audits.
  • Large gaps remain in meeting traditional audit expectations in AI system integration.
  • Access to necessary information and resources significantly influences audit design.

Evidence Used

  • 4,259 documents scanned for the review

Methods / Approach

  • Reflexive thematic analysis

Limitations Stated by the Authors

  • The study may not have captured all relevant AI audits due to the limited number of documents scanned.

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The interpretation and summary provided here are based on a textual analysis of the source. Some nuances and complexities may be lost in this process.
  • The study does not provide specific examples of the audits reviewed, making it difficult to fully understand their scope and methodology.

Open Questions

  • What are the specific audit practices that can effectively capture failures across various levels in AI system integration?
  • How can the AI community prioritize system integration as a core strategy for addressing AI risk?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The study does not provide details on what constitutes traditional audit expectations in AI auditing.

Raw Abstract / Excerpt

As AI systems become increasingly integrated into diverse interfaces and applications, model-centric audits are insufficient to address risks arising from interactions among system components and deployment environments. System integration has long been central to software audits in safety-critical domains such as aerospace. However, its role in AI auditing remains underexplored. Scanning through 4,259 documents, we present a scoping review of AI audits that treat system integration as a core tenet of evaluation (n = 58). Using reflexive thematic analysis, we analyze their elements, actors, enablers, and constraints. We find that the corpus represents an emerging yet still fragmented form of AI auditing: few existing measures target integration-specific risks; large gaps remain in meeting traditional audit expectations; and access to necessary information and resources significantly influences audit design. Nonetheless, integration can be categorized across three sites (inter-component, system-environment, and multi-system), each serving the functions of risk exploration, risk determination, coordination, and procedural regularity. Deviating from other types of evaluations, these audits assess qualities specific to system integration, including compatibility, completeness, and oversight. This review calls on the AI community to prioritize system integration as a core strategy for addressing AI risk, and to develop audit practices capable of capturing failures across components, environments, and systems beyond the reach of component-level evaluation.