regulated-ai

Agility infrastructure for regulated AI. Trace, Policy, Evidence under EU jurisdiction. Apache 2.0, on-premise capable.

Behavioral Trust Clustering a thermodynamic governance layer that reduces LLM hallucination by 52% on HumanEval. Drop-in wrapper for any decoder. MIT.

Governance patterns for autonomous AI agents in regulated financial services: DEFCON state machine, Sovereign Veto, Audit Chain, EU AI Act mapping

Model-neutral Agent IAM runtime for regulated AI workflows: short-lived capabilities, API-body enforcement, credential brokering, and tamper-evident audit trails.

kolm — the AI compiler. Compile any task into a signed .kolm artifact that runs locally. RS-1 open spec, MIT.

Secure CI/CD and infrastructure patterns for regulated AI/ML deployments

A modular architecture for AI-powered regulated investigation. 8 universal base components · 7 plug-and-play clusters · 6 deployment profiles.

Interpretability-first ML pipeline wrapper for regulated industries

Compliance and audit middleware for LLM agents operating in regulated financial workflows.

Policy enforcement for AI agents in regulated environments (FERPA, HIPAA, GLBA, GDPR): framework adapters for CrewAI, AutoGen, LangChain, Semantic Kernel, Haystack

A starter repository for documenting, reviewing, and releasing regulated AI systems.

Reference architecture for wrapping a frontier model so it can ship in a regulated workflow: gateway, permission-aware retrieval, two-sided guardrails, identity/KYB-KYC, pre-registered eval gate, tamper-evident audit. The model is the easy part; the wrap is the product.

Reference architecture for accountable AI document workflows with RAG-style retrieval, validation, human review, and audit logging.

Regulated multi-agent operations platform with governance, review, and audit-friendly workflows.

Fiducial Mesh — open specification (CC-BY-4.0) for cradle-to-grave traceability of AI-assisted engineering: sovereign agent identity, policy, audit, and knowledge-grounding primitives for regulated, certifiable AI. SPEC-001 + HDBK-001.

Traceable, human-controlled impact assessment for clinical-trial protocol amendments, with evidence, audit trails, and fail-closed escalation.

AI compliance framework for human oversight - EU AI Act Article 14 HITL/HOTL gate - 人类监督闸门（Human Oversight Gate）（SynomosAI 四支柱 · human-oversight-gate）

Evidence-producing evaluation harness for AI under regulatory constraint: requirement traceability, calibration gate, sealed evidence bundle

​A behavioral specification for AI agents in high-compliance environments. Uses semantic primitives to constrain probabilistic LLMs with deterministic guardrails while preserving human accountability.

Cryptographic receipts for agent traces: hash chain, Ed25519, JCS canonical.