Bigfoot Analytics audits AI systems in financial services against EU AI Act requirements — before regulators do.
Download Credit Risk Audit Download Fraud Detection Audit Book a Free CallReal findings from real audits — not theoretical checklists.
Absent checking account data ranked as the single most influential feature in a credit scoring model — above loan duration, credit amount, and age. The model was approving applicants because their data was missing, not because they were creditworthy.
HIGH SEVERITYA model predicting "legitimate" for every transaction achieves 99.83% accuracy. Our audit found the model only captured 45.2% of fraud value in euros — meaning thousands of euros in fraud escaped in the test period alone.
CRITICALPCA anonymisation applied for GDPR compliance made 93.3% of model features legally unexplainable. Customers have a right to explanation under GDPR Article 22 — the anonymisation makes that right impossible to fulfil.
HIGH SEVERITYYoung adults under 25 faced systematically lower approval rates and an 18-percentage-point accuracy gap compared to older applicants — not because of higher credit risk, but because of structurally shorter credit histories.
HIGH SEVERITYDefault threshold of 0.5 left significant fraud undetected. Our threshold analysis showed that changing to 0.3 raises F1 from 0.79 to 0.85 with near-zero increase in false positives — a simple fix with major compliance impact.
HIGH SEVERITYNeither audited system had documented override procedures, monitoring dashboards, or shutdown protocols — all mandatory under Article 14. Human oversight was assumed, not architected.
CRITICALTechnical compliance auditing for financial AI systems — not checkbox questionnaires.
We test your AI model against EU AI Act requirements and deliver a structured findings report with severity ratings, article mappings, and concrete remediation actions. Covers Articles 6, 9, 10, 12, 13, 14 and 15.
Quarterly re-audits and drift detection to keep your model compliant as the law evolves and your model changes over time. Automated alerts when compliance thresholds are breached.
Not sure where to start? Book a free 30-minute EU AI Act readiness call. We assess your current exposure and tell you exactly what needs to be done before August 2026.
Our methodology is not a consulting framework. It is derived from published academic research.
Eight Compliance-by-Design principles derived from systematic review of 12 peer-reviewed sources from ABDC A and A* ranked journals. Introduces the Dynamic Compliance Architecture — a framework in which regulatory compliance is native, continuous, and automated.
Read on SSRN → Download PDFBigfoot Analytics was founded by Sanket Shrivas — a finance professional with banking operations experience including a Citi Bank New York project, now specialising in EU AI Act compliance for financial AI systems.
Currently completing a Master's in Data Science at IU International University Berlin. Independent researcher with published work on AI auditing design principles available on SSRN.
The name reflects what we do — we find things others miss. The compliance gaps hiding in plain sight inside your AI systems. The findings that standard questionnaire tools cannot detect because they never actually test the model.
Based in Berlin, Germany. Serving financial institutions and fintechs across the EU.
Book a free 30-minute compliance readiness call. No commitment. Just clarity on where you stand.
contact@bigfoot-analytics.com Connect on LinkedIn