HN Showcase
Back to browse
GitHub Repository

The flight recorder for AI agents. Tamper-proof, ISO 42001 + HIPAA-compliant audit trails for everything AI agents do on your system. File changes + shell commands + PHI detection + agent identity.

7 starsPython

Unworldly – A flight recorder for AI agents (tamper-proof, HIPAA)

by dilawargopang·Feb 25, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerSolve My ProblemZero to OneDark Horse

GAP research proved models hide harmful actions in tool calls—this catches them at runtime.

Strengths
  • •Backed by peer-reviewed arXiv paper (GAP benchmark) proving the problem exists across 6 frontier models
  • •Deterministic allow/deny/redact with no LLM-in-loop means 55μs overhead and no hallucination risk
  • •Tamper-proof SHA-256 hash chain + ISO 42001 compliance addresses real regulatory need
Weaknesses
  • •Tied to AI agent safety niche; won't matter for teams not running autonomous agents yet
  • •HIPAA/ISO 42001 claims need third-party audit—currently self-certified
Category
Security
Target Audience

AI agent operators, compliance-focused teams, healthcare/finance using autonomous AI

Similar To

Anthropic Workbench · OpenAI Agents SDK governance · LangChain callbacks

Similar Projects

Security●●Solid

Air – Open-source black box for AI agents (tamper-evident audit trails)

Instead of another observability dashboard, this project builds a provable audit trail: an OpenAI-compatible reverse proxy that vaults prompts in MinIO and links calls with an HMAC-SHA256 tamper-evident chain, plus replay tooling (replayctl) and Jaeger traces. The cryptographic audit chain and the one-line SDK wrap are clever and practical; the real operational work left to teams will be key management and storage/retention strategy.

WizardryNiche Gem
shotwellj
213mo ago
Security●●●Banger

Korveo – a local firewall for AI agents

Local firewall blocks agent credential leaks before they leave your machine.

Zero to OneSolve My Problem
amitbidlan
1316d ago