HN Showcase
Back to browse
I tested 11 AI frameworks for basic security – none passed

I tested 11 AI frameworks for basic security – none passed

by Acacian·Apr 2, 2026·1 point·1 comment
Visit ProjectView on HN

AI Analysis

●●●BangerBig BrainSolve My Problem

Streaming guardrail catches semantic PII that regex misses — based on real LangChain issues.

Strengths
  • •Semantic PII detection using AI, not just regex patterns.
  • •2.65ms overhead across 4 guardrail scans — 0.5% of typical LLM latency.
  • •Auto-instrumentation adds security with one function call, no behavior changes.
Weaknesses
  • •AI agent security is emerging — Lakera Guard and PromptArmor already compete here.
  • •Browser playground demo doesn't prove production reliability at scale.
Category
Security
Target Audience

Developers building AI agents with LangChain, CrewAI, or OpenAI

Similar To

Lakera Guard · PromptArmor · Guardrails AI

Similar Projects

Security●●Solid

Ziran, security testing for AI agents

Instead of testing LLMs in isolation, ZIRAN models agents as stateful, tool‑wielding actors — it flags dangerous tool chains (e.g. read_file → http_request), runs multi‑phase trust‑building attack campaigns, and surfaces attack paths in a knowledge graph. The A2A protocol support and remote HTTPS scanner with protocol auto‑detection are concrete, unusual features; the repo (docs, examples, PyPI) suggests solid follow-through, though usefulness will depend on adapter coverage for real deployments.

Big BrainNiche Gem
leoneperdigao
104mo ago
Security●●Solid

Khaos – Every AI agent I tested broke in under 30 seconds

Auto-patching LLM calls to inject faults and log telemetry is a neat technical trick that lets you fuzz real agent runs without changing your stack. The repo ships six intentionally vulnerable example agents and a CLI (discover/run/ci) with eval packs for security and resilience, so you can reproduce attacks and gate releases. It feels like an early, practical toolkit that fills a gap in agent security testing — adoption and more community-playbooks will determine how far it goes.

Big BrainNiche Gem
exordex
114mo ago
Security●●Solid

AgentArmor – open-source 8-layer security framework for AI agents

Eight-layer defense-in-depth for AI agents when Guardrails AI only handles inputs.

Solve My ProblemShip It
AgastyaTodi
1063mo ago