sipi.bot · compliance

EU AI Act compliance for agent spend governance

The EU AI Act classifies AI systems in financial decision-making as high-risk. Autonomous agents authorizing payments trigger governance requirements under Articles 9-14.

What EU AI Act requires for agent spend

Under the EU AI Act (effective 2026), agent-triggered payments trigger Article 9 risk management, Article 10 data governance, Article 11 technical documentation, and Article 14 human oversight. A deterministic pre-spend firewall with human-in-the-loop escalation satisfies the governance baseline.

How sipi.bot maps to EU AI Act controls

Requirementsipi.bot capabilityEvidence
Documented policy gateevaluate_spend in <5ms against rulesPolicy version per decision
Immutable audit logTamper-evident decision logFull JSON audit trail
Human escalationFLAGGED → approval queueQueue history with reviewer ID
Access controlAgent-key-scoped policiesAuth + authorization log
Compliance starts with the pre-spend check. The single most important control is deterministic evaluation before the transaction commits. See plans →

Frequently asked questions

What does EU AI Act require for agent spend?

A documented, enforced policy gate on every autonomous transaction, with an immutable audit trail and human escalation. sipi.bot evaluates every transaction against your rules in <5ms, logs the decision with policy version, and queues flagged transactions for human review.

Is sipi.bot certified for this?

sipi.bot is not a certifying body. It provides the technical controls — deterministic evaluation, audit logging, human escalation — that satisfy these governance requirements. Your auditor maps these to the framework criteria.

How do I produce audit evidence?

The sipi.bot audit log exports every decision (transaction, amount, merchant, rule triggers, verdict, policy version, timestamp) as structured JSON. This log is tamper-evident and can be fed into SOC 2, ISO 27001, or GDPR audit trails.

Stop runaway agent spend before it happens

sipi.bot is a pre-spend firewall for autonomous AI agents — approve, block, or flag every transaction in under 5ms.

See plans →