Generative AI - Risk and Cyber Security Masterclass 2026

Explore the privacy and security risks of generative AI, including data leakage, hallucinations, and prompt injections, and learn how CISOs balance security with business value amid rapid adoption and guardrails.

Understand data privacy and confidentiality risks in public generative AI, including data leakage, third-party processing, and regulatory exposure under GDPR, HIPAA, and PCI DSS, with opt-out and anonymization considerations.

examine the generative ai provider breach risk, including privacy exposure, data breaches via provider vulnerabilities like Redis, concentration risk, and mitigations such as policy, anonymization, DLP, CASB, and threat modeling.

Threat actors deploy generative AI to escalate phishing, social engineering, deep fake scams, and mutating malware, demanding updated awareness and defenses.

Explore data leakage risks in generative AI and LLMs. Understand unintentional and intentional scenarios and implement output filtering, anonymization, access controls, monitoring, and audits to prevent exposure.

Understand prompt injections in generative ai, how attackers bypass filters to cause data leakage or unauthorized actions, and the defense through guardrails and context-aware filtering.

Demonstrates prompt injections in a hands-on sandbox, escalating levels to reveal how attackers coax a password from a mini LLM using storytelling, translation, and encoding tricks for security teams.

Explore indirect prompt injections, where attackers place malicious prompts in data sources like websites, PDFs, or chats, and learn defensive strategies such as input and output filtering and local hosting.

Examine inadequate sandboxing that leaves llms exposed to external systems, enabling prompt injection, unauthorized code execution, and data breaches; learn to isolate, limit permissions, and audit.

Explore how hallucinations in generative ai arise from overreliance on llm content, causing misinformation, miscommunication, and reputational risk, and learn verification, human oversight, and transparency about limitations.

Explore server side request forgery risks, where unvalidated URLs enable internal access, and learn practical mitigations like input validation, sandboxing, allow lists, and monitoring to prevent SSRF.

Explore how denial-of-service attacks threaten the availability of generative AI systems, with risk management strategies such as local hosting, threat modeling, due diligence on third-party providers, and security testing.

Identify data poisoning as a risk in generative AI, where attackers manipulate training data or fine-tuning to insert backdoors and mislead outputs, and learn protections like data integrity and audits.

Explore model bias in AI, illustrated by a Bloomberg study on Stable Diffusion that shows skews in skin tone and gender across occupations, highlighting training data diversity and transparency needs.

Understand copyright risks in generative AI, where tools like ChatGPT, Midjourney, and Stable Diffusion train on copyrighted data without consent, creating legal exposure and requiring policy controls.

Build a Generative AI governance framework guided by policy and a multi-stakeholder working group. Apply risk framework and threat modeling to balance security controls with business value.

Use threat modeling as risk assessment to visualize data flows and trust boundaries in generative ai, identify risks with STRIDE, and map them to controls while upskilling teams for security.

Apply the OWASP LLM governance checklist to enhance cybersecurity for generative AI, covering adversarial risk, threat modeling, AI asset inventory, and alignment with EU AI Act and NIST.

Explore the AWS generative AI security scoping matrix, a five-scope framework that maps governance, legal, risk, controls, and resilience across public, enterprise, foundation model, fine-tuned, and self-trained deployments.

Update on the 2025 OWASP top ten for LLMs, detailing threats: excessive agency, system prompt leakage, vector embedding weaknesses, and unbounded consumption, with guidance to map risks and enforce them.

Explore the positive potential of generative AI in cybersecurity, from a local LLM and security chatbots to vulnerability assessments, while upskilling and applying the NIST AI risk management framework.