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

Vertex AI

Vertex AI is Google Cloud's managed platform for AI model development, deployment, and operations.

Key facts

• The Hacker News reported a Vertex AI SDK flaw involving model-upload bucket handling, with potential bucket-squatting risk during upload flows ^[src-181].
• The source is relevant to AI Watch because agentic and ML systems often inherit cloud SDK security assumptions rather than treating storage paths as part of the attack surface ^[src-181].

Related

• Google Cloud
• AI Supply Chain Security
• Agent Security

Historical recovered snapshot — 2026-08-03

Preserved from pr-123. Documentary recovery on 2026-09-09; claims below retain their historical source scope and were not freshly verified. Current synthesis above remains authoritative.

Vertex AI's evaluation layer supports adaptive and static rubrics, deterministic metrics, custom Python functions, generic pointwise/pairwise judges and tool-use checks ^{[src-265][src-266]}. The third-party notebook demonstrates a common evaluation path for external APIs through LiteLLM, Model Garden MaaS and independently served BYOM endpoints ^[src-268].