Model Online / On-Site
Duration
Modules 5 modules

Course Objective & Scope

In regulated sectors such as finance, healthcare, telecoms and the public sector, sending customer records or trade secrets in cleartext to cloud LLM providers creates major compliance barriers under KVKK, GDPR and industry regulations. This course covers end-to-end privacy-preserving architectures that embed privacy engineering across the full LLM lifecycle—ingestion, RAG, prompt dispatch, inference and logging.

Participants implement PII masking with Microsoft Presidio and Turkish NER models, reversible pseudonymisation, data residency routing between regional cloud and local open-weight models, and confidential computing options to produce a privacy architecture for their organisation.

Who Should Attend?

  • AI and software architects designing LLM and RAG solutions in regulated industries
  • Privacy engineers and data security specialists
  • Engineering teams responsible for cloud and platform infrastructure
  • Prerequisite: knowledge of Python, enterprise API architectures and core data protection concepts

Key Takeaways

  • Design hybrid routing architectures between local and cloud LLMs based on data classification
  • Build PII masking and reversible pseudonymisation flows supporting both Turkish and English text
  • Apply pre-indexing anonymisation and access controls in RAG knowledge bases
  • Configure regional cloud data residency, zero data retention and customer-managed keys (CMK)
  • Assemble a technical evidence package for KVKK cross-border transfer rules and DPIAs

Curriculum

Privacy Threats and the Data Lifecycle in LLM Systems
  • Personal data exposure points across prompts, context, embeddings, fine-tuning, caches and logs
  • Technical constraints imposed on LLM data flows by KVKK, GDPR and sector regulations
PII Detection, Masking and Reversible Pseudonymisation
  • Microsoft Presidio, regex/checksums (national IDs, IBANs, payment cards) and Turkish NER models
  • Assigning placeholders via a token vault and safely detokenising model responses
Anonymisation in Structured and Unstructured Data
  • k-anonymity, l-diversity, generalisation and suppressing quasi-identifiers
  • Reducing contextual re-identification risk in RAG documents and training/eval datasets
Data Residency and Sovereign / Hybrid Model Deployment
  • Region pinning, Zero Data Retention (ZDR) and abuse-monitoring exemptions in cloud providers
  • Designing a privacy-aware router that directs sensitive prompts to on-prem open-weight models (vLLM)
Encryption, Embedding Privacy and Confidential Computing
  • Vector encryption and tenant isolation against embedding inversion attacks
  • Confidential GPU computing (TEEs) and an end-to-end privacy architecture design workshop

Other Courses in This Category

  • Building PII Detection and Masking Gateways for LLMs

    Hands-on training on building a low-latency PII detection, masking and de-pseudonymisation gateway in front of LLM APIs using Microsoft Presidio, GLiNER and Turkish rule/NER recognisers.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Data Anonymization and Synthetic Test Data Generation for AI

    Teaches how to replace production customer data in testing, analytics and LLM evaluation environments using k-anonymity, differential privacy and SDV/LLM-based synthetic data generation.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Data Minimization and KVKK Compliance in AI Logging and Observability

    Teaches how to scrub personal data from LLM and agent telemetry in OpenTelemetry, Langfuse and SIEM logs, and engineer retention policies, access controls and KVKK compliance into observability pipelines.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Agent Goal Hijacking: Detection and Defense in Autonomous AI

    Covers Agent Goal Hijack—the top OWASP Agentic AI risk—showing how attackers divert autonomous planning loops, persistent memory and multi-step execution, and how to defend against it with intent verification.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Agentic AI Supply Chain and Model Context Protocol (MCP) Security

    Covers tool poisoning, rug pulls, tool shadowing and supply chain risks in Model Context Protocol (MCP) servers, third-party agent tools and A2A integrations, and how to build a secure MCP architecture.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Preventing Unexpected Code Execution in LLM and Agent Systems

    Teaches how to prevent remote code execution (RCE) and injection vulnerabilities caused by executing LLM- and agent-generated Python, SQL, shell or template outputs using AST validation and safe output handling.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View

Get in touch