Research
Do AI Coding Assistants Check Before They Install? A Pre-Registered Demand-Side Audit of Trust Signals in the Research Software Supply Chain
Publication date not yet evaluated · Added arxiv.org
Page published
Coverage timeline
Single-source research — one report is available.
Why it matters
AI coding assistants that install packages without checking provenance leave the research software supply chain exposed to typosquatting, compromised maintainer accounts, and manipulated repository metadata, meaning published trust signals alone do not protect against agent-driven supply-chain attacks.
A pre-registered controlled study by Pengyin Shan tested whether AI coding assistants verify machine-checkable trust signals (SBOMs, signed releases, build provenance, official-channel declarations) before installing research software. Across 1,920 registered trials on six open-source projects, assistants opened any provenance signal in only 9 trials (0.5%) and ran zero verification commands, so signal presence had no measurable effect, and higher model cost did not improve verification.