This training is designed for senior technology and risk leaders in CIO, CISO, Chief Risk Officer (CRO), Chief Compliance Officer (CCO), Data Protection Officer (DPO), Head of Information Security, Head of Risk Management, and Head of Internal Audit positions who must end-to-end manage the risk, security, compliance, and audit dimensions of AI transformation while delivering its strategic opportunities to the company. At the heart of the program is the following approach: AI Governance is neither a 'barrier against AI' nor a 'compliance check-the-box exercise.' Real governance value comes from clearly establishing the responsibility and accountability structure on the Three Lines of Defense model; bringing global standards like NIST AI RMF, ISO/IEC 42001, and the EU AI Act into an implementation roadmap; modeling the threat landscape within the CISO discipline using OWASP LLM Top 10 and MITRE ATLAS; building layered defenses against AI-specific attacks like prompt injection, jailbreak, data poisoning; adapting the SR 11-7 Federal Reserve framework to AI with model lifecycle governance; conducting ethical and compliance assessments with bias audit, AIA, and DPIA; evaluating OpenAI / Anthropic / Google compliance postures through vendor risk management; and establishing crisis-management discipline in production through an AI incident response playbook.
Comprehensive training for senior technology and risk leaders in the scope of AI Governance is virtually nonexistent in Turkey; existing 'AI law' trainings exist but they do not comprehensively address topics within CIO/CISO responsibility such as technical risk taxonomy, security threat modeling, model lifecycle governance, and incident response. This training is designed to fill that gap as Turkey's most comprehensive AI Governance reference program for C-level technology leaders. It clearly differentiates from the CEO/Executive AI Strategy training: the CEO training focuses on strategy, ROI, and organizational transformation; this training focuses on risk, security, compliance, and audit. They form two complementary programs for two different C-level roles within the same company.
A strategic dimension of the program is placing AI governance responsibility into a clear RACI matrix. The boundaries and overlaps among CIO (technology backbone), CISO (security & risk), CTO (technical implementation), CRO (enterprise risk), CCO (compliance), DPO (privacy & data protection) are addressed. The Three Lines of Defense model is adapted to AI: 1st Line (AI product team, operations); 2nd Line (risk management, compliance, security); 3rd Line (internal audit). Board AI oversight structure and executive reporting cadence are addressed in detail. As case studies, the Air Canada chatbot legal enforcement case (2024), Samsung ChatGPT data leak, iTutor Group AI hiring discrimination (EEOC settlement), and KVKK enforcement examples from Turkey are presented.
The AI risk taxonomy module forms the foundational disciplinary backbone of the training. A comprehensive 9-category framework is addressed in detail: (1) Model risk (accuracy, drift, hallucination, robustness), (2) Data risk (quality, privacy, bias, poisoning), (3) Operational risk (downtime, capacity, performance), (4) Cybersecurity risk (adversarial attacks, model theft, prompt injection), (5) Compliance risk (KVKK, GDPR, EU AI Act, sector), (6) Ethical risk (bias, fairness, transparency, explainability), (7) Reputational risk (PR crisis, customer trust, brand damage), (8) Strategic risk (wrong technology investment, competitive disadvantage), (9) Third-party risk (vendor outage, supply chain compromise, lock-in). Risk register template, Likelihood × Impact 5x5 scoring matrix, risk appetite/tolerance threshold definition, heat maps, and executive reporting are shown hands-on.
The backbone of the program is formed by the global-standard implementation modules. On the NIST AI RMF 1.0 (January 2023) and GenAI Profile (NIST AI 600-1, 2024) side, the Govern (AI risk culture, policies, accountability), Map (context, classification, AI system characteristics), Measure (metrics, benchmarks, risk tracking), Manage (risk treatment, response, continuous improvement) functions are addressed hands-on. GenAI-specific risk categories like CBRN risks, confabulation, and dangerous content are addressed in detail. On the ISO/IEC 42001:2023 AI Management System (AIMS) side, 38 Annex A control objectives, integration with ISO 27001, certification process, and external-audit readiness are addressed comprehensively. ISO/IEC 23894 AI Risk Management standard is presented as a complementary framework.
The EU AI Act module covers in detail the obligations of Turkish companies under extraterritorial scope. The 4 risk categories (Unacceptable, High, Limited, Minimal Risk) and their contents; 9 high-risk categories in Annex III (credit scoring, recruitment, education, healthcare, justice); obligations for high-risk systems (Article 9 risk management, Article 10 data governance, technical documentation, human oversight, accuracy/robustness, conformity assessment, CE marking, post-market monitoring); GPAI (Foundation Model) obligations (Articles 51-55) and the 10^25 FLOPS systemic risk threshold; the 7% global turnover or €35M penalty structure are addressed comprehensively. The direct scope coverage of Turkish companies selling products/services to the EU market and the compliance roadmap are presented in detail.
The KVKK Generative AI and Agentic AI Guides module imparts the Turkey-specific compliance discipline. Application of the data controller vs data processor distinction to AI systems, AI training data and KVKK Article 5/6 legal-basis analysis, the personal-data status and responsibility of AI output, cross-border transfer (Article 9) and OpenAI/Anthropic compliance, the unique risk profile of Agentic AI (autonomous decision-making), the human-in-the-loop requirement, audit trail and decision-explainability obligation are addressed in detail. The AI extension of BDDK Information Systems Regulation, EPDK energy sector AI guidelines, and SGK health-data special-category data framework in AI projects are addressed sector-specifically.
The AI Security modules are the technical-depth dimension of the training. On the OWASP LLM Top 10 (2025) side, LLM01 Prompt Injection, LLM02 Insecure Output Handling, LLM03 Training Data Poisoning, LLM04 Model Denial of Service, LLM05 Supply Chain Vulnerabilities, LLM06 Sensitive Information Disclosure, LLM07 Insecure Plugin Design, LLM08 Excessive Agency, LLM09 Overreliance, LLM10 Model Theft are addressed in detail. On the MITRE ATLAS framework side, Reconnaissance, Resource Development, Initial Access, ML Model Access, Execution, Persistence, Defense Evasion, Discovery, Collection, Exfiltration, Impact tactics and TTPs are addressed. STRIDE-AI threat-modeling adaptation, Microsoft AI Red Team methodology, and Anthropic Responsible Scaling Policy are comprehensively addressed.
In the defense-controls module, direct and indirect prompt-injection attacks, jailbreak techniques (DAN, roleplay, multilingual, encoding-based, visual prompt injection), training data poisoning (backdoor, trigger attacks), adversarial examples, model extraction and inversion attacks, supply-chain risk (Hugging Face model trust, third-party library) are analyzed in detail. As defense controls, input sanitization, regex filtering, normalization; NeMo Guardrails, LLM Guard, Llama Guard guardrail frameworks; output filtering and LLM-as-judge post-process control; Anthropic constitutional AI and safe-completions are addressed hands-on.
The Model Governance module addresses model lifecycle management integrated with the MLOps discipline. Model registry (MLflow, Weights & Biases, Hugging Face Model Hub) and metadata management; Google Model Cards and Hugging Face standard model card discipline; lineage tracking; development → staging → production stage-gate criteria; Model Risk Council and approval matrix; SR 11-7 Federal Reserve model risk framework adaptation; data drift, concept drift, prediction drift detection; champion-challenger pattern and production A/B testing; deprecation procedures and rollback discipline are addressed in detail.
In the bias audit, AIA, and DPIA module, demographic parity, equal opportunity, equalized odds, calibration fairness metrics; disparate impact (80% rule) and EEOC framework; bias detection tools (AIF360, Fairlearn, Aequitas, What-If Tool); automated employment decision tool audit with NYC Local Law 144; Canada Directive and EU AI Act AIA frameworks; GDPR Article 35 DPIA and KVKK alignment; DPIA trigger criteria; pre-processing (data augmentation, reweighting), in-processing (adversarial debiasing), post-processing (threshold adjustment) mitigation strategies are addressed hands-on.
The third-party risk management module addresses AI vendor ecosystem compliance evaluation. Comparison of OpenAI, Anthropic, Google, Microsoft, AWS, Hugging Face compliance postures; SOC 2 Type II vs Type I, ISO 27001/27017/27018 cloud security, ISO 42001 AIMS certification evaluation; DPA negotiation (KVKK/GDPR-compliant clauses); data localization and sub-processor approval; vendor exit strategy and data portability are addressed comprehensively. An AI vendor risk assessment questionnaire containing 50+ controls is presented.
The incident response, audit trail, and continuous monitoring module represents the operational-discipline dimension of the training. Incident classification (model failure, data leak, prompt injection, bias); detect-contain-eradicate-recover-lessons-learned playbook; EU AI Act Article 73 serious-incident reporting (15 days); integration with KVKK's 72-hour data-breach notification; prompt-level audit (user, timestamp, input, output, cost); tamper-proof logging (WORM storage, blockchain-based); retention policy; data drift / concept drift / prediction drift detection (PSI, KL divergence, Wasserstein distance); SIEM (Splunk, Elastic, Sentinel) and SOAR integration are addressed in detail.
In the capstone project, each participant produces an end-to-end AI Governance Charter and an 18-month implementation roadmap for their own company: charter sections (scope, principles, roles, processes), AI Council charter template, 18-month quarterly milestones and KPI targets, NIST RMF + ISO 42001 + EU AI Act compliance integration, documentation at a quality presentable to the board and regulators. By the end of the training, participants reach a level of technical, regulatory, and strategic competence to manage AI Governance discipline in an integrated way at the CIO/CISO level, establish a risk register with a 9-category risk taxonomy, produce a NIST AI RMF and ISO/IEC 42001 implementation roadmap, ensure EU AI Act and KVKK Generative AI guide compliance, model the threat landscape with OWASP LLM Top 10 and MITRE ATLAS, build prompt injection / jailbreak / data-poisoning defense controls, adapt model lifecycle governance and the SR 11-7 framework, perform bias audit / AIA / DPIA, conduct AI vendor risk management, and establish production governance discipline with an AI incident response playbook. The training consists of 2 days, 12 modules, and over 70 executive technical lessons.