DSG

DSG ONE

ProofGate Control Plane

Production Evidence

DSG Gateway evidence: 6/6 passed, 95% rubric score

DSG is a deterministic AI Action Governance Gateway for high-risk tool execution. It places a verified control layer between AI intent and production systems, checking policy, entitlement, risk, approval, and mathematical invariants before an action can execute.

Benchmark verdict

PASS

Pass rate

100%

Checks passed

6 / 6

Average latency

2047 ms

Min latency

1029 ms

Max latency

4254 ms

Comparison rubric

190 / 200

Comparison percent

95%

SMT2 invariants

6 / 6

Deterministic governance claims

Deterministic control

DSG uses deterministic control logic for high-risk tool calls, reducing reliance on probabilistic model judgment at the execution boundary.

Action-layer hallucination risk reduction

DSG is designed to reduce hallucination-driven actions by preventing AI outputs from directly executing tools unless they satisfy policy and invariant checks.

Mathematical invariants

DSG enforces runtime safety invariants over organization identity, actor identity, registered tool/action matching, approval requirements, entitlement checks, and evidence writability.

Audit-proof state transitions

DSG turns AI tool execution into an auditable state transition with request hashes, record hashes, and exportable evidence.

Verified benchmark checks

Connector registration
Gateway custom HTTP execution
Monitor Mode plan-check
Monitor Mode audit commit
Audit events API
Audit export API

Public vendor-baseline coverage

DSG is production-tested. The listed market leaders are public documentation baselines only; this page does not claim vendor runtime benchmark results.

Zapier
n8n
Make
Workato
Temporal

Evidence boundary

This page reports an internal production benchmark run for marketplace and demo validation. DSG does not claim to make language models hallucination-free. It controls whether AI-proposed actions can affect business systems by requiring deterministic policy and invariant checks at the execution boundary. Cost reduction claims are design goals and risk-reduction mechanisms unless supported by customer production studies.

npm run benchmark:evidence
npm run benchmark:vendors:baseline

Production Gateway Result:
pass: true
total: 6
passed: 6
failed: 0
passRate: 100%
avgLatencyMs: 2047
minLatencyMs: 1029
maxLatencyMs: 4254

Comparison Rubric:
score: 190 / 200
percent: 95%

SMT2 Runtime Invariants:
pass: true
checks: 6 / 6
passRate: 100%

Public Vendor Baseline:
pass: true
dsgRuntimeTested: true
publicDocVendors: 5
vendorRuntimeTested: 0