Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification.

AI Supply Chain Security

AI supply chain security covers model artifacts, training and upload pipelines, storage destinations, credentials, provenance, and deployment handoffs. The Vertex AI SDK bucket-squatting report is a reminder that model upload paths can be part of the attack surface [src-181].

Related

  • Agent Security
  • AI Cybersecurity Governance
  • Google Cloud

Historical recovered snapshot — 2026-08-18

Preserved from ai-watch-src329-335-okf-replacement-20260818t150256z. Documentary recovery on 2026-09-09; claims below retain their historical source scope and were not freshly verified. Current synthesis above remains authoritative.

  • Wiz’s Snowflake proof of concept shows how untrusted issue metadata can become shell code inside CI workflows and expose adjacent service credentials, even after automated security review [src-332].
  • The LiteLLM incident shows a cross-project credential cascade: compromised Trivy artefacts executed in CI, stole a PyPI publishing token, and enabled direct upload of malicious LiteLLM wheels using Python startup hooks [src-333][src-334].
  • Vendor estimates of 2,488 domains or 2,500+ organisations describe potential exposure, not confirmed compromise. The 434,000-pipeline headline appears to conflate 433,909 files with 118,829 CI-runner dumps [src-333][src-334].
  • [src-332] Gal Nagli / Wiz — Snowflake CI/CD injection and Red Agent proof of concept (2026-08-17)
  • [src-333] CloudSEK — LiteLLM supply-chain exposure analysis (2026-08-11)
  • [src-334] Hudson Rock — LiteLLM leaked-corpus analysis (2026-08-12)

Evidence checked on 14 September 2026

Wiz reports MCP authentication bypass, post-authentication code execution and exposed administration across internet-facing LiteLLM deployments. Its scan found 9.6% of 3,074 instances used default or absent authentication. [src-474] Affected paths have different prerequisites. Wiz reports patches for assigned vulnerabilities; the authenticated pass-through abuse case is a separate design exposure.

Robin Cartier perspective

This page is part of Robin Cartier's working AI knowledge graph: a practical research layer for production AI, recommendation systems, experimentation, GEO, and agentic web readiness.

The useful next step is to connect this concept back to applied product leadership and operating models.

Recommended next

Keep reading from this thread

From 500 indexed pages and articles.

  1. Wiki concept AI Cybersecurity Governance Covers the policies and operational controls needed to secure AI systems, including model supply chains, cloud SDKs, agent permissions, storage Related by chain
  2. Wiki concept Vertex AI Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification. Related by litellm
  3. Wiki concept Agent Security Focuses on constraining, observing, and recovering AI agents that can use tools, credentials, cloud resources, or production workflows. Related by security