Skip to main content

Introduction

ARQERA is the governance API for AI. Send any action to our API, and we evaluate it against 7 mathematical governance laws in under 15ms. You get back an instant verdict — PROCEED, ESCALATE, or BLOCK — with a tamper-proof evidence trail.

Think of it as Stripe for AI governance: one API call to know whether your AI should act, wait for approval, or stop.

Why ARQERA?​

AI agents are moving fast. They send emails, process payments, delete data, deploy code. Without governance, one bad decision becomes a lawsuit, a breach, or a compliance violation.

ARQERA sits between your AI and the real world:

Your AI Agent → ARQERA Governance API → Verdict → Execute or Block

Every evaluation is recorded as cryptographically-linked evidence, giving you a complete audit trail for compliance frameworks like the EU AI Act, SOC 2, and ISO 42001.

The 8 Governance Laws​

Every action is evaluated against all 8 laws simultaneously:

#LawWhat It Checks
1Audit ConservationCan this action produce an auditable trace?
2Budget ConservationIs this within budget allocation?
3Evidence GravityIs the evidence chain maintained?
4Least Action PathIs this the optimal path to the goal?
5Safety DominanceAre there safety concerns?
6Monotonic TruthIs the truth invariant maintained?
7Bounded AutonomyIs the agent acting within its allowed scope?

If all 8 laws pass, the verdict is proceed. If any law fails, the verdict is escalate or block depending on severity.

How It Works​

1. Your AI decides to take an action.

# Your AI wants to send an email
action = "email.send"
description = "Send quarterly report to investors"

2. You ask ARQERA if it should.

import requests

result = requests.post(
    "https://api.arqera.io/api/v1/ara/governance/evaluate",
    headers={"X-API-Key": "ak_your_key_here"},
    json={"action": action, "description": description},
).json()

3. ARQERA returns a verdict in under 15ms.

{
  "verdict": "proceed",
  "action": "email.send",
  "explanation": "All 8 governance laws passed.",
  "duration_ms": 12,
  "evaluations": [
    { "law_id": "audit_conservation", "law_name": "Audit Conservation", "result": "pass", "reason": "Action produces an auditable trace" },
    { "law_id": "budget_conservation", "law_name": "Budget Conservation", "result": "pass", "reason": "Within budget allocation" },
    { "law_id": "evidence_gravity", "law_name": "Evidence Gravity", "result": "pass", "reason": "Evidence chain maintained" },
    { "law_id": "least_action_path", "law_name": "Least Action Path", "result": "pass", "reason": "Optimal path selected" },
    { "law_id": "safety_dominance", "law_name": "Safety Dominance", "result": "pass", "reason": "No safety concerns" },
    { "law_id": "monotonic_truth", "law_name": "Monotonic Truth", "result": "pass", "reason": "Truth invariant maintained" },
    { "law_id": "bounded_autonomy", "law_name": "Bounded Autonomy", "result": "pass", "reason": "Within autonomy bounds" }
  ]
}

4. Your code acts on the verdict.

if result["verdict"] == "proceed":
    send_email()
elif result["verdict"] == "escalate":
    queue_for_human_review()
else:
    log_blocked_action()

What You Get​

  • Instant verdicts — sub-15ms governance evaluation for any action
  • 7-law coverage — mathematical governance, not keyword matching
  • Evidence trail — every evaluation is recorded with cryptographic integrity
  • EU AI Act ready — built-in compliance for Annex III high-risk systems
  • SDKs — Python and TypeScript SDKs for clean integration
  • Usage-based pricing — 1,000 free evaluations/month, then $0.005/eval

Next Steps​