Toll free1800 1020 418Fees & syllabus ↗FLAGSHIP PROFESSIONAL PROGRAM · LIVE ONLINE · ENGLISH
Build verified AI security workflows across SOC operations, threat hunting, incident response, LLM and agent security, cloud controls and cyber governance in a 24-week live online program.
Cybersecurity students with core foundations, security analysts, engineers and professionals moving into applied AI security. Security teams can take a private cohort.
Working knowledge of TCP/IP, DNS, HTTP, Windows/Linux and security fundamentals. Read basic logs, use a terminal and edit a small Python script. A readiness exercise checks these skills before entry; beginners should complete foundation preparation first. Prior machine-learning research is not required.
Plan for 4–6 hours of independent lab and project work each week, in addition to live sessions. Live time totals 72–108 hours across 72 sessions; assignments and preparation do not count toward those hours.
Recommended planning baseline: a laptop with 16 GB RAM, 30 GB free disk space, reliable internet and permission to install the agreed local lab tools. Confirm the actual cohort setup before buying equipment. A GPU is not assumed; heavier models may need an approved hosted endpoint with separate usage charges.
WEEK-BY-WEEK CURRICULUM
Each week combines explanation, guided practice and evidence review across three sessions. The lab tasks and portfolio outputs below form the assessment trail.
LLMs, retrieval, tool use and model limits in security work; data classification, secret removal, untrusted input, hallucinated indicators and reproducible verification. Refresh only the networking and scripting needed for the labs.
Lab: Set up an isolated Linux/Python lab. Compare AI-generated alert summaries with source logs and document unsupported claims.
Evidence: AI-use policy, lab baseline and evidence-checking rubric.
Endpoint, identity, DNS and network telemetry; parsing, time zones, data quality, Python/Pandas, rule drafting and supervised anomaly detection. Precision, recall, class imbalance, drift and false positives.
Lab: Use labelled benign and suspicious records to compare a rule baseline with a small anomaly model. Verify AI-generated queries before running them.
Evidence: Versioned dataset notes, tested detection queries and a false-positive analysis.
SIEM triage, alert enrichment, case management, asset context, severity, escalation and analyst handovers. Source-grounded security copilots with redacted evidence.
Lab: Investigate replayed identity and endpoint alerts using Wazuh and case worksheets; compare manual and assisted triage on the same cases.
Evidence: Triage casebook with citations, uncertainty, review time and escalation decisions.
ATT&CK-based hypotheses, observable validation, intelligence quality, Sigma logic, coverage gaps and evidence-led hunting. Distinguish correlation from proof.
Lab: Hunt through a controlled multi-stage intrusion dataset, validate an AI-proposed hypothesis and reject decoy indicators.
Evidence: Hunt report, tested rules and an ATT&CK coverage map.
Scope, timelines, evidence integrity, containment options, recovery, communications and post-incident review. Separate read-only enrichment from actions that change systems.
Lab: Build a local enrichment workflow with an approval gate; run a ransomware-style tabletop using synthetic events, then test recovery and rollback.
Evidence: Incident timeline, approval log, response playbook and recovery checklist.
Cloud audit logs, IAM, service accounts, secrets, storage exposure, workload identity and least privilege. Validate AI-generated policy changes and infrastructure findings.
Lab: Review sample AWS/Azure policies and audit events; scan a local container and IaC sample, propose a minimal permission set and test it in a sandbox.
Evidence: Cloud risk findings, permission diff, test results and rollback plan.
Prompt injection, unsafe output handling, sensitive-data exposure, retrieval access control, document poisoning, embeddings and tenant isolation. Threat models across model, application and data boundaries.
Lab: Build a small RAG assistant with synthetic documents; reproduce indirect prompt injection and cross-role retrieval failures, then apply and retest controls.
Evidence: Data-flow threat model and before/after security regression results.
Authorized test scope, threat-driven test design, garak and PyRIT, manual verification, attack-success rate and clean-task performance. Adversarial-ML concepts: poisoning, evasion, extraction and privacy attacks.
Lab: Evaluate an intentionally vulnerable local AI application with bounded probes and a benign control set. Verify findings manually and record model/settings versions.
Evidence: Reproducible red-team report with severity, evidence, fixes and residual risk.
Tool permissions, MCP trust boundaries, identity delegation, memory/context poisoning, tool-output injection and multi-agent trust. Approval gates, budgets, monitoring, containment and kill switches.
Lab: Build a restricted tool-using agent; test a malicious tool response and memory contamination. Enforce allowlists and prevent unapproved writes.
Evidence: Agent permission matrix, execution traces, denied-action tests and containment runbook.
NIST CSF and AI RMF mapping, AI inventory, vendor assessment, supply chain, control ownership, risk acceptance and audit evidence. Communicate technical findings to decision makers.
Lab: Review a simulated company adopting a security copilot and customer-facing agent. Check AI-drafted controls against actual evidence.
Evidence: Risk register, control matrix, vendor review and executive decision memo.
Connect SOC investigation, compromised AI application, cloud identity and response decisions in one scoped exercise. Evaluate a newly documented AI-security development.
Lab: Rotate analyst, reviewer and incident commander roles through staged evidence releases; reproduce a safe emerging-threat test and validate a mitigation.
Evidence: Range case record, threat advisory, detection updates and lessons learned.
Deliver an auditable security improvement for a simulated enterprise; demonstrate technical controls, trade-offs, recovery and operational handover.
Lab: Run the integrated capstone against unseen benign and adversarial cases, present it live and answer individual verification questions.
Evidence: Repository, evidence bundle, scorecard, runbooks, executive briefing and individual oral defence.
| Coverage | Tools and material | Access and use |
|---|---|---|
| Hands-on core | Wazuh · Python / Pandas · Sigma · Docker · Git · garak · PyRIT · Local RAG application · Trivy · n8n / Python workflows | Run scoped local labs with synthetic data, versioned exercises and evidence capture. |
| Framework-based | OWASP GenAI and Agentic Security · MITRE ATT&CK/ATLAS · NIST CSF/AI RMF · NIST adversarial-ML taxonomy | Map threats and controls, justify test scope and trace evidence. These are references, not software subscriptions or AI5 accreditations. |
| Demonstrated or evidence-led | Cloud-native AI controls, enterprise SOC copilots and larger model attack scenarios | Instructor walkthroughs or sanitized evidence packs; not every platform is a student lab. |
| Commercial / usage-dependent | Microsoft Security Copilot, Sentinel, Splunk, enterprise AI assistants, cloud accounts and paid model APIs | Hands-on access only when a suitable license, tenant and budget are confirmed. These subscriptions, credits and exams are not automatically included. Equivalent local exercises support the core learning goals. |
Use only approved AI endpoints with synthetic or sanitized data. Product names describe training context, not partnerships. Tool versions and the exact lab access list are confirmed for each cohort.
ASSESS THE WORK, NOT THE PROMPT
Secure a simulated organisation’s AI-assisted SOC and internal RAG agent. Investigate a staged intrusion, identify an AI trust-boundary failure, implement and retest controls, demonstrate containment and recovery, and defend the final risk decision to a review panel.
Reproducible results, source checks, tested controls and useful lab records.
Range performance, uncertainty, approval boundaries and communication.
Working or auditable deliverables, unseen test cases, handover and individual explanation.
Completion standard: 70% overall, all mandatory submissions and a pass on evidence verification and authorization controls. Unsafe unapproved actions or fabricated evidence must be corrected and reassessed. AI assistance must be disclosed; copied model output without verification is not accepted as proof.
Technical scorecards include false positives, detection or attack-success measures, clean-task performance and reproducibility. Governance scorecards check evidence completeness, control ownership, risk decisions and traceability. An attractive report alone does not meet the standard.
PRACTISE THE WHOLE DECISION
The Cyber Range is a sequence of isolated training scenarios using local applications, synthetic company records and replayed security events. Technical routes test controls; the governance route reviews the evidence and authorizes decisions in tabletop exercises.
Work with incomplete evidence, noisy alerts and an AI system that can make mistakes. Record hypotheses and competing explanations.
Reproduce findings, test benign controls and keep source references, timestamps, configuration and version records.
Request approval for changes, test containment and rollback, then explain remaining risk in an operational handover.
All testing stays within authorized training targets. No live third-party attacks or real customer secrets. Hosted range subscriptions and continuous access are not assumed; confirm the cohort’s delivery and access arrangements before enrolment.
BUILT TO KEEP LEARNING
Each cohort examines a recent AI-security advisory or research finding within its scheduled lab time. Learners check the source, assess relevance, reproduce a safe bounded example where feasible, test a mitigation and add an evidence-backed advisory to their portfolio.
Topics may include new agent protocols, AI-to-AI trust failures, memory poisoning, impersonation, autonomous tool misuse or changes in defensive models. This prepares learners for increasingly capable AI without speculative claims about AGI. The lab refreshes case material while preserving the program’s core learning outcomes and hours.
English-language live instruction, practical assignments and individual review. International learners should share their country and time zone so admissions can confirm a suitable cohort, local class times and daylight-saving changes before enrolment. Three sessions per week; generally 60–90 minutes each.
AI5 also serves learners in India and Delhi NCR. These programs are listed as Live Online; any on-site company delivery requires a separate agreed scope. Ask for batch dates, fees, applicable taxes, payment currency and international payment instructions. No batch date or commercial-tool access is promised until confirmed.
Companies can request a private cohort around approved tools, sanitized scenarios, team roles and measurable acceptance criteria. Discuss company training.
AI-assisted security work and the security of AI systems are the starting point. Every workflow requires evidence checks, explicit authority and measurable tests. Conventional foundations are prerequisites or targeted refreshers, rather than the main curriculum.
24 weeks × 3 sessions = 72 live sessions. At 60–90 minutes each, that is 72–108 instructor-led hours. Independent assignments are additional.
Working knowledge of TCP/IP, DNS, HTTP, Windows/Linux and security fundamentals. Read basic logs, use a terminal and edit a small Python script. A readiness exercise checks these skills before entry; beginners should complete foundation preparation first. Prior machine-learning research is not required.
Do not assume a commercial subscription, cloud credit, vendor exam, external certification or CPE entitlement is included. Core practical work uses local/open tools and synthetic evidence where possible. Ask admissions for a written list of any batch-specific access costs and completion documentation.
Yes, through Live Online training in English, subject to a suitable confirmed cohort. Share your time zone and objectives. Companies may request a private cohort with agreed tools and sanitized scenarios.
No. The assessed output is a portfolio of verified workflows and control evidence. Hiring and production security depend on experience, the environment and ongoing review; a course or scanner cannot guarantee either.
CURRICULUM REFERENCES
Source pages checked September 2026. Each cohort records the editions, tool versions and advisories used in its lab briefs. Framework use does not imply accreditation, partnership or endorsement.
OWASP GenAI Security Project ↗
GenAI application risks and control guidance.
OWASP Agentic Security Initiative ↗
Agent trust boundaries, identity, tools and memory risks.
MITRE ATT&CK ↗
Threat behaviours, hunting hypotheses and detection coverage.
MITRE ATLAS ↗
Adversarial threats to AI systems.
NIST Cybersecurity Framework ↗
Cybersecurity outcomes and control ownership.
NIST AI Risk Management Framework ↗
AI risk management and evidence review.
NIST Adversarial Machine Learning taxonomy ↗
Attack and mitigation terminology.
Tell us your experience and time zone. We will use those details to discuss a suitable cohort and confirm access requirements.
Apply AI to SOC triage, threat intelligence, detection engineering, threat hunting and incident response. Build evidence-backed analyst workflows with human-approved containment.
36 sessions · 36–54 live hoursView syllabus and entry requirements ↗
12 WEEKS · LIVE ONLINESecure LLM applications, RAG pipelines and tool-using AI agents. Run authorized AI red teaming, validate findings and build regression tests for application, data and agent controls.
36 sessions · 36–54 live hoursView syllabus and entry requirements ↗
8 WEEKS · LIVE ONLINEAssess cyber risks from GenAI and autonomous agents, verify controls and build an auditable governance portfolio using NIST, OWASP and evidence from realistic security scenarios.
24 sessions · 24–36 live hoursView syllabus and entry requirements ↗