A stream-based runtime-verification framework for generating hard real-time C code.
Controlled PyTorch experiments testing xLSTM-style streaming forecasting and runtime assurance under telemetry faults, distribution shift and guarded adaptation.
A runtime assurance kernel for LLM-based systems: graded verdicts, hashed commitments, bit-identical replay.
Read-only deterministic anomaly detection for mission-critical infrastructure using variance, allocation velocity, and amplification signals.
Trusted autonomy T&E runtime that links mission needs, hazards, scenarios, telemetry, evidence, verification reports, and hash-chained ledgers so AI/autonomous decisions can be reviewed instead of merely trusted.
Source-available governed AI cognition architecture for bounded episodic execution, reality-corrected world modeling, evidence-gated skill synthesis, transfer challenge, replay validation, and human review. No AGI claims.
Vendor-neutral control family (ARA-1..8) for deterministic, tamper-evident, independently-verifiable runtime assurance of AI agents. Proposed for NIST CAISI / OWASP GenAI / ISO SC42.
Governed cognition substrate for AI engineering agents: mission envelopes, evidence-bound belief/plan graphs, quarantined memory, model-role review, sentinel checks, and BlackFox handoffs under human authority.
Governed AI engineering control plane: federated workload identity, delegated least privilege, pre-dispatch MCP/API authorization, human approval, revocation, and cryptographically signed authority receipts with tamper/replay/rollback detection and independent verification. AI proposes. Humans decide.
A layered architecture for running autonomous AI agents under deterministic safety constraints — a trusted, hard-coded harness that validates and bounds the output of a probabilistic model.
Geometric reachability-based safety filtering for air traffic control
Mission-assurance control plane for spacecraft edge autonomy: governs when onboard AI may act, abstain, degrade, retain data, request ground review, or fall back, with signed evidence, deterministic replay & traceability from mission requirements to decisions.
Physical causal evidence for autonomous fault diagnosis — open-world mechanism hypotheses, safe discriminating tests, independently adjudicated recovery. Authority-separated, receipted.
Deterministic postcondition verification for AI agents and autonomous systems using independent observed state.
Source-available, measurement-first pulsed-energy testbed for tri-sector storage/discharge control, derated storage, phase authority, sensor-truth checks, energy accounting, kill criteria, evidence bundles, and human-reviewed scale-up gates.
Curated research, tools, benchmarks, standards, and open-source systems for assuring autonomous systems.
A governed world-model evidence layer for AI agents: simulate bounded scenarios, track assumptions, score prediction-vs-reality error, and produce human-reviewable execution evidence.
Runtime behavioral assurance and drift detection for AI agents. Monitors CloudTrail/Bedrock traces. Detects control violations and policy breaches. Maps to NIST AI RMF, ISO 42001.
Open Verification Kernel (OVK) is an open-source, solver-agnostic verification layer for AI-agent engineering workflows.
Signal-Time-Authority (STA): a pre-commitment controllability framework for runtime AI oversight.
To associate your repository with the runtime-assurance topic, visit your repo's landing page and select "manage topics."