Explainability mandate
Every automated fraud decision must be explainable on request. Why a transaction was declined. Which signals were weighted. In language a regulator — or a customer — can read.
SENTR · EU AI ACT
EU AI Act enforcement begins August 2026. Your automated fraud decisioning is affected.
From August 2026, the EU AI Act requires any payment operator using automated decision-making systems — including fraud detection — to maintain an explainable AI audit trail. Fines for non-compliance: up to 7% of global annual turnover.
⏱ August 2026 — 3 months from now
Up to 7% of global annual turnover in fines for non-compliance.
What EU AI Act requires for fraud detection
Four obligations. All of them apply to automated fraud decisioning.
Audit trail requirement
A complete timestamped record of every decision: inputs, model version, outputs, reviewer identity where human review occurred. Auditable on demand. No reconstruction from memory.
Fines: up to 7% global turnover
Non-compliance with high-risk AI system obligations carries fines of up to 7% of global annual turnover. Lesser violations: up to €20M or 4% of turnover.
August 2026 — enforcement begins
The deadline is fixed. Grace periods do not apply to high-risk system classifications. Fraud decisioning systems are in scope. Build your compliance baseline before, not during, an audit.
All regulatory references sourced to EU AI Act (Regulation 2024/1689).
Where mid-market fraud ops are exposed
Five gaps between your current compliance posture and August 2026 requirements.
- HIGH RISK
Decision audit trails
You don't have timestamped decision logs tied to individual transactions. Regulators will ask for them. Manual reconstruction from logs and analyst memory is not an audit trail — it is a liability.
- HIGH RISK
Model explainability
Your fraud model returns a score. It cannot articulate why it declined a specific transaction in terms a regulator or customer can read. Black-box decisioning is the primary compliance exposure under Article 13.
- MEDIUM RISK
Rule and threshold change management
No formal audit trail when your detection rules or risk thresholds change. No versioning of decision logic. A regulator asking "what were your model parameters on 15 March?" has no clean answer.
- MEDIUM RISK
False positive liability
You're blocking legitimate customers. You are not measuring the financial or compliance impact. EU AI Act requires operators to demonstrate that high-risk automated decisions are not causing disproportionate harm to individuals.
- MEDIUM RISK
Regulatory readiness
No playbook for producing audit reports on demand. No compliance software. When an enforcement notice arrives, the answer is weeks of manual preparation — not a 45-minute export.
Mid-market teams carry the same EU AI Act obligations as Tier-1 banks — without the 200-person infrastructure those banks built to meet them.
How SENTR addresses each exposure
A compliance baseline built during the proof period — before you commit to anything.
SENTR shadow pathway — five stages from read-only evaluation to production decisioning. Each step compounds explainability and audit evidence; outcomes align with what regulators expect under the EU AI Act (Reg. 2024/1689).
Read-only connection
SENTR connects to your live transaction stream via read-only API. Your existing platform unchanged. Nothing in production is touched.
Baseline capture · evidence chain begins
Parallel processing
SENTR processes every transaction you process. Applies our decisioning logic. Writes zero decisions to your production system.
Shadow proof · no customer impact
Explainability log generated
Every transaction decision generates a human-readable log: which signals fired, confidence score, decision rationale. Stored, timestamped, exportable.
Article 13–style rationale · per decision
Intelligence report delivered
Full fraud intelligence report + compliance gap assessment against EU AI Act requirements. Regulator-ready baseline from Day 0.
Gap map vs. enforcement timeline
Production deployment
Move SENTR to live decisioning. Continue building your explainability baseline in production. Audit trail compounds from go-live.
Live trail · compounding audit evidence
SENTR generates a full explainability log for every fraud decision — human-readable, regulator-ready, exportable in under 45 minutes. Shadow mode builds that baseline during the proof period. You do not need to be live on SENTR before August. You need to be able to demonstrate you are building toward compliance. Shadow mode starts that clock.
FREE DOWNLOAD
Get the 27-point Compliance Checklist
We built a 27-point compliance checklist for risk and compliance teams at growth-stage PSPs and iGaming operators. Traffic light scoring. 90-day roadmap on page 9.
- 27-point compliance audit — every EU AI Act obligation mapped
- Traffic light scoring — red / amber / green by obligation
- 90-day implementation roadmap — page 9
- EU AI Act section-by-section mapping
- Free. No credit card. PDF delivered to your email.
Download the checklist
See how shadow mode builds your audit trail
Book a 20-minute Architecture Session. We map your current fraud stack, walk you through what your EU AI Act exposure looks like, and show you what shadow mode would build during a 50-day evaluation.
Book your Architecture Session →
No pitch deck. No contract. If shadow mode is not the right fit for your current setup, we will tell you on the call.