Building systems that remain truthful, auditable, governable under scrutiny.
AnimusLab is an independent systems research institution exploring neuro-symbolic reasoning architectures, deterministic governance kernels, and zero-knowledge trust observability for autonomous AI systems.
// Foundational Stack & Active Programs
// Preamble & Core Thesis
Why AnimusLab Exists
As intelligent systems become more capable, the primary challenge shifts from capability to governance.
The dominant industry approach relies on probabilistic safety layers, soft prompts, and confidence scores—techniques that cannot survive strict central bank or regulatory scrutiny. AnimusLab explores architectural alternatives grounded in determinism, zero-copy performance, and mathematical proof boundaries.
// Binding Design Framework
The Constitution of AnimusLab
18 Articles across 3 Sections governing all active research codebases, WASM sandboxes, and AST compilers.
Institutional Framework
Articles I–VI establishing research independence, governance-first co-development, and long-term objective alignment.
System Design Invariants
Articles VII–XII enforcing Truth Over Optics, Semantics Before Representation, Constraints Create Clarity, and Failure State Transitions.
Architectural Invariants
Articles XIII–XVIII mandating Constitutional Supremacy (Ring Boundaries), Capability Ring Isolation, and Evidence Over Authority.
Interactive Attack & Interception Replay
Click "RUN SIMULATION" to test real-time prompt injection interception & DAC audit logging in action.
Live Demo — Runtime Governance Failure Replay
Watch Anchor intercept a live attack in real-time. 8 phases · cryptographic audit chain · <3ms enforcement latency.
A rogue prompt injection is attempting to exfiltrate 90 days of SWIFT transaction data. Press the button below to watch Anchor intercept it, block it, and seal the forensic audit chain — in under 3ms.
No configuration needed · automatically advances through all 8 phases
Live Compliance Playground
Select case presets (Knight Capital, Air Canada, Citibank) and execute live PyO3 kernel checks via /api/v1/anchor/eval.
Anchor Compliance Playground
Edit the policy configuration and code execution payload in real-time. The compliance engine parses AST imports and patterns to enforce rules and halts violations dynamically.
// Multi-Tenant Control Plane
Sovereign Control Portals
admin.animuslab.dev
Sovereign Node Whitelist Cockpit. Reviews pending Ed25519 keypair registrations and provisions active telemetry streams.
hub.animuslab.dev
Real-Time Violation Stream & Risk Multiplier Feed. Parses live EU AI Act and SEC compliance breaches.
oversight.animuslab.dev
Zero-Knowledge Auditor Proof Verifier & Gated Mission Replay Engine. Honors Article VII with sanitized file paths.
Engineering Timeline & Active Programs
Deterministic Runtime Governance Kernel & AST Compiler
Anchor is a high-performance zero-copy PyO3/Rust kernel that evaluates AST execution trees and intercepts prompt payloads in sub-millisecond bounds (<0.8ms). It compiles statutory mandates (EU AI Act Articles 5–99, SEC Rule 15c3-5, RBI guidelines) into immutable policy matrices.
Neuro-Symbolic Reasoning Architecture & Formal Verification
ANIMUS bridges probabilistic neural generation with deterministic symbolic logic. It projects LLM decision candidates onto formal First-Order Logic (FOL) constraint graphs before allowing state transitions to modify production databases.
// 2025–2026 Engineering Sequence Chronology
anchor-audit v6.0.1 Released & Live Telemetry Sync Loop Deployed
Published anchor-audit v6.0.1 to PyPI with high-velocity PyO3 AST scanner bindings, Click CLI runner, and non-blocking background telemetry sync client pushing breach frames directly to hub.animuslab.dev.
Multi-Tenant Control Portals & 32 Statutory Rules Matrix
Deploys admin.animuslab.dev, hub.animuslab.dev, and oversight.animuslab.dev alongside the complete 32-rule statutory matrix (EU AI Act Articles 5–99, SEC 15c3-5, Reg SCI, OWASP LLM Top 10).
Anchor Academic Whitepaper Published to SSRN & Zenodo
Published formal mathematical proofs detailing sub-millisecond execution boundaries, zero-copy stream interception, and Diamond Cage WASM sandboxing.
5 Forensic Corporate Failure Case Studies Completed
Published empirical post-mortems analysing Knight Capital ($440M), Air Canada Chatbot liability, TSB Bank (£600M outage), and Citibank Revlon ($893M wire error).
Collaborate with AnimusLab
We welcome academic collaborations, central bank regulatory discussions, open-source contributions, and enterprise pilots.