Guardrails & safety — page 6

Cheap fail-open semantic edge layer for Jev (TypeSafe System One): decision service + confidence policy + decision logging + blind eval harness + skill router over 1,000+ skills. Stdlib-only Python.

Benchmarking TypeSafe's Jev against four rival prompt-injection detectors on 11,900 labelled prompts

An LLM built without a language model — using TypeSafe Jev's non-generative AI for contextual response selection. Zero hallucination. Zero prompt injection.

JEV Prompt Injection Guardian is a prompt injection quarantine and risk-scoring system for LLMs powered by JEV (jev-1.13.0), TypeSafe AI's innovative "System One" model, as its primary engine. By eliminating ambiguous, long-winded natural language explanations, it instantly computes rigorously calibrated probability risk-scores.

Spec-compliant universal agent skill: fast TypeSafe Jev semantic judgments (intent detection, code ranking, quality gates, safety checks) via a bundled stdlib-only CLI

Autonomous settlement circuit breaker protecting ledgers against CVE/RCE balance bypasses, nocturnal draining, and prompt injection attacks using TypeSafe Jev.

SemanticPolicy adds testable semantic decisions to .NET applications: rules a decision model answers, measured on labelled examples, for business logic and AI agents, with no lock-in to one provider.

Production Agent Decision Harness for TypeSafe Jev & System One models — confidence-gated policy routing, shadow mode, and trajectory verification

Backs the WordPress AI plugin's Comment Moderation with TypeSafe Jev, through Connector for TypeSafe Jev

LLM-based permission gate for OpenCode agents: Jev auto-approves routine tool calls and fails open to a human prompt on anything risky or uncertain.

Jev Policy Engine is the first universal Policy-as-Code SDK built for TypeSafe AI's Jev System One model. It allows RevOps, DevOps, and Security teams to define strict, deterministic AI governance rules in YAML, and execute them at high speed.

TypeSafe Jev plays Battleship: one ~100-question typed fan-out per turn returns a calibrated hit-probability heatmap that is also the move policy.

TypeSafe jev-powered LLM model router: classify → policy → tiered pools → OpenAI-compatible passthrough proxy (streaming, tools, full agent loops)

Reproducing the real $1 Chevy Tahoe chatbot exploit and testing 6 AI defenses (incl. TypeSafe's real Jev API) against 36 prompt-injection attacks.

Fail-closed Jev policy gate for Hermes Agent tool calls

Live-stream encoder settings from telemetry and history: TypeSafe's Jev decision model inside a deterministic safety envelope. Deno, Node, edge runtimes.

A visual control room for typed AI evaluations, structured decision-making, and confidence-aware policy gating using Vercel AI SDK and TypeSafe Jev.

Which shell commands may your coding agent run without asking? An allowlist proves what it can; Jev, a no-text model, judges only the rest. Claude Code hook + CLI, measured.

TypeSafe AI (Jev System One) plugin for OpenClaw - sub-100ms group triage, tool safety guardrails, compaction curation, and model routing

Shell command safety gate for the Pi coding agent, backed by TypeSafe's Jev judgment model

UiPath bring-your-own-guardrail connector backed by the TypeSafe Jev System One model: plain-language agent policies enforced as calibrated probabilities.

A runtime assurance layer for tool-using AI agents: it observes an execution agent independently, combines hard safeguards with Agentic Assurance, and makes the final control decision through a deterministic policy.

ASSAY-001: independent, pre-registered verification of TypeSafe Jev's calibration and type-safety claims. Split verdict, published in full.

A local lab for seeing what TypeSafe's Jev judgment model does on realistic business cases: typed answers, probabilities, policy in code, receipts.