EXIN AI Security Professional study materials
EXIN AI Security Professional study resources

EXIN AI Security Professional case studies

EXIN AI Security Professional · Put your knowledge to work in context.

The EXIN AI Security Professional (AISP) is a new, independently examined certification from EXIN, a certification body with over 40 years of experience. AISP is built on the OWASP AI Exchange and is aligned with EU AI Act Article 15 and Article 4, the EU AI Act security standard (prEN 18282), and reflected in ISO/IEC 27090. The EXIN AI Security Professional case studies are 5 cases and about 3,526 words.

1

Indirect Prompt Injection in a Retail Support Agent

Nordwind Retail runs a customer-support assistant on a commercial LLM. It ingests uploaded PDFs through OCR, retrieves policy passages from a vector store, and can call order status, refund, and address change APIs.

738 words, about 4 min
2

Label Poisoning in a Fraud Model Supply Chain

LumenPay is a licensed e-money institution using a gradient-boosted fraud model for card-not-present transactions. A third-party vendor supplies 14 merchant-risk features, including chargeback history and domain age, through weekly unsigned S3 drops.

701 words, about 4 min
3

Recruitment AI Facing the EU AI Act

TalentSift BV, based in Utrecht, licenses a CV-screening tool to 300 EU employers. The model ranks applicants by predicted job fit using historical hiring outcomes; employers make the final decision.

705 words, about 4 min
4

Fine-Tuning an Open-Weight Model Under the GPAI Rules

Alicia Labs, a Spanish startup, fine-tunes an open-weight 70B model on 2.1 trillion tokens of licensed Spanish legal text and releases the weights as 'Lex-70B' under its own name. It estimates 8×10^24 FLOPs for the fine-tune; the base model provider reported 3×10^25 FLOPs.

691 words, about 3 min
5

Model Extraction Through a Forecasting API

Helio Predict sells API access to a proprietary demand-forecasting model used by 90 retailers. Over 11 days, one 'analyst' account made 2.4 million queries with systematically varied price, promotion, weather, and holiday inputs on a dense grid; 92% came from 14 IP addresses in two countries and volume was 40× baseline.

691 words, about 3 min