Deploying Safe and Secure AI Agents
About This Course
The rapid adoption of Large Language Models (LLMs) and agentic AI systems by organizations introduces a critical challenge: innovating without exposing themselves to unacceptable risks such as data leakage, hallucinations, prompt injection, and compliance breaches.
As AI systems gain capabilities, attackers’ incentives increase, and the consequences of misuse escalate. The exponential growth of LLMs powers new capabilities but also brings risks of malicious use (e.g., scams, cyberattacks, misinformation), malfunction (e.g., bias, unsuitable deployment), and systemic failure (e.g., privacy violations, copyright issues).
This course directly addresses these urgent needs by equipping professionals with a foundational and practical understanding of agentic AI security architecture, hands-on experience with guardrail technologies and threat detection, and best practices for secure design, deployment, and compliance.
Learners will explore multi-layered guardrails to mitigate risks such as hallucinations, prompt injection, and compliance breaches. Practical application is emphasized through labs using industry tools. The course prepares participants to deliver trustworthy, production-ready AI solutions through real-world deployment, validation, and continuous monitoring.
What You'll Learn
- Identify and analyse agentic AI risks, distinguishing safety versus security failure modes (e.g., jailbreaks, indirect prompt injection, data leakage) to enable scoping of effective mitigations.
- Apply evaluation frameworks and continuous monitoring so that enterprise deployments meet measurable gates pre-release and detect drift or abuse post-release in production contexts.
- Design defence mechanisms for agentic AI by proposing viable input/output filters, policy checks, and tool-use constraints that reduce exploitable risk surfaces in production.
- Introduce and evaluate the concept of guardrails for security, compliance, and governance in AI-in-the-loop systems, operationalising policy and auditability across the lifecycle.
This course is targeted at below job roles:
- Software developers
- DevSecOps developers
- Solution Architects
- IT Managers
Entry Requirements
Familiar with Python programming (basic scripting, packages, Jupyter Notebook)
Basic machine-learning knowledge (datasets, training, and inference)