CAIS case studies
CAIS · Put your knowledge to work in context.
CAIS - often expanded as the Certificate of AI Security and associated with the Cloud Security Alliance - is positioned as an AI security credential for practitioners who must govern, build, and defend machine-learning systems rather than conventional cloud infrastructure alone. One important caveat frames everything on this page: the CSA education page for CAIS returned a 404 Not Found error when retrieved. The CAIS case studies are 5 cases and about 3,678 words.
1Tiering a Vendor Chatbot Under a Half-Built AI Policy
Meridian Regional Bank is 10 days from launching a vendor-hosted LLM support assistant for 40 contact-centre agents. The assistant reads customer profile fields (name, address, last four card digits, account status) from the CRM and drafts replies; it cannot move funds.
783 words, about 4 min2Indirect Prompt Injection in a Prior-Authorisation Agent
Northvale Health Plan uses an LLM agent to draft prior-authorisation decisions. The agent reads faxed PDFs uploaded by providers, extracts clinical criteria, queries an internal policy database, and emails a draft to a nurse reviewer.
720 words, about 4 min3Regurgitation in a Fine-Tuned Support Model
Larkspur Retail fine-tuned an 8-billion-parameter open-weight model on 14 months of customer-service chat logs - 4.2 million conversations - to power a support assistant in three markets. Redaction before training used regular expressions for email addresses and phone numbers only.
716 words, about 4 min4A Poisoned Model Artifact in the MLOps Pipeline
Corvus Pay is a 400-person fintech with 60 data scientists. Its model pipeline builds containers in GitHub Actions, installs dependencies with pip install -r requirements.txt with no hashes or pins, and downloads model checkpoints from a public hub by repository name only.
737 words, about 4 min5Drift, Fairness, and a Regulator Letter
Halcyon Insurance runs a fraud-scoring model over 1.8 million claims a year. Its monitoring dashboard tracks uptime, latency, and error rate.
722 words, about 4 min