Audit of 15 self-hosted AI stacks finds 14 ship with zero NetworkPolicies
No-Peanut-6988 · reddit · 2026-09-27
A security team audited default Helm/Docker configs for 15 popular AI stacks (LiteLLM, vLLM, Ray, Weaviate) and found serious defaults: LiteLLM's migration Job embeds the Postgres password in plaintext; KubeRay accepts unauthenticated job submissions while workers have passwordless sudo (instant root); some Kubernetes MCP servers get cluster-wide Secret read and pod exec with no auth; vector DB charts ship with anonymous read/write. 14 of 15 charts have zero NetworkPolicies and default token automounting enabled. Remediation snippets and open-source probe scripts are available.
More from Infra
- India's Sarvam launches Token Factory, hosting top open models on local infra — HarveenChadha · 2026-11-16
- Gemini 3.5 Flash now runs on air-gapped Google Distributed Cloud, offline up to 12 months — shashib · 2026-11-16
- Gemma.c: full Gemma 1/2/3 multimodal inference in a single C file, ~72 tok/s on CPU at INT8 — This-Purchase-3325 · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16