Skip to content

AI Cyber Security Foundations

Secure AI with the
skills to govern, test,
and protect it.

A practical seven-course program for the people responsible for securing AI, from governance and risk to application security, architecture, and red teaming.

See the Courses See the Price
7Courses
13Hours of Study
3Exam Attempts
1Price
  1. Foundations AI Security Foundations
  2. Governance Governing AI Adoption
  3. Prompt Injection Prompt Injection
  4. Agentic AI Agentic AI and Excessive Agency
  5. Data Security AI Data Security
  6. Assurance Assurance
  7. Red Teaming Red-Teaming AI Systems

The Seven Courses

Taken in order, because each builds only on the ones before it. Every course is a written module with worked examples and prompts you can use the same day, a 20-question scenario exam graded on our server, and a deliverable your security team keeps.

  1. Course 1 · 90 minutes

    AI Security Foundations: What Is Actually New

    What about securing an AI system is genuinely new, and what is ordinary application security under a new name. The threat model every later course builds on, in the vocabulary the rest of the industry uses: the Open Worldwide Application Security Project (OWASP) Top 10 for large language model applications, and MITRE's Adversarial Threat Landscape for Artificial-Intelligence Systems (ATLAS).

    You leave with an AI threat model for one feature, alongside the security review your team already writes.

  2. Course 2 · 90 minutes

    Governing AI Adoption: The Register and the Gate

    Most companies govern the AI they buy for staff and nothing about the AI they build and ship to customers, which is the half that carries their name. This course builds what a regulator, an auditor, an enterprise customer and your own board are actually asking for.

    You leave with a register of every AI system the company runs, and the company-wide AI use policy that governs it.

  3. Course 3 · 120 minutes

    Prompt Injection, Direct and Indirect

    Why prompt injection is a property of the design rather than a bug to patch, why filtering input does not resolve it, and how to contain what a successful injection can reach. The course that retires "we sanitize user input" as an answer.

    You leave with a probe set built from one feature's own inputs, a measured result for each probe, and the review of what a successful one would have reached.

  4. Course 4 · 120 minutes

    Agentic AI and Excessive Agency

    An agent's security posture is what it can do in one run before anyone notices, and almost nobody has written that number down. Autonomy, tool use, blast radius, and where a human in the loop is worth having.

    You leave with the authority statement for one agent: every tool it holds, the bound on each, what one run can do unattended, and how it is stopped.

  5. Course 5 · 120 minutes

    AI Data Security: Retrieval, Embeddings and What Leaks

    Retrieval makes copies of your data in places your data controls do not reach, and an embedding is one of those copies. Tenant isolation, disclosure through retrieval, and deleting what was copied.

    You leave with a map of every retrieval source: its classification, where access is enforced, where copies end up, and how each copy is deleted.

  6. Course 6 · 120 minutes

    Assurance: Frameworks, Regulation and Vendor Risk

    The National Institute of Standards and Technology AI Risk Management Framework, the international AI management standard ISO/IEC 42001, the European Union's AI Act, and how to assess an AI vendor. Almost every assurance question comes down to the register, the threat models and the deletion paths, so this course arranges the earlier work for someone outside the company.

    You leave with the assurance pack an auditor, a regulator or an enterprise customer asks for, and one assessment per AI vendor.

  7. Course 7 · 120 minutes New

    Red-Teaming AI Systems: Scope, Attack and Report

    A red team against an AI system is scoped by what the system can reach, not by how the model can be fooled. The track turns around and attacks what the first six built: the application, the model and its supply chain, and the infrastructure it runs on, with every finding measured as a rate over repeated trials and graded by what it reached. It teaches the method and how to build a practice range on your own staging systems; it does not replace time in a hands-on lab.

    You leave with a red-team report: the scoped plan, every finding with its reproduction rate and severity, and the retest.

All Seven Courses. One Price.

There is no price per course. The seven are sequenced so each depends on what comes before it.

One price, every course

AI Cyber Security Foundations

  1. AI Security Foundations: What Is Actually New
  2. Governing AI Adoption: The Register and the Gate
  3. Prompt Injection, Direct and Indirect
  4. Agentic AI and Excessive Agency
  5. AI Data Security: Retrieval, Embeddings and What Leaks
  6. Assurance: Frameworks, Regulation and Vendor Risk
  7. Red-Teaming AI Systems: Scope, Attack and Report
  • All seven courses, about 13 hours of study
  • Every course’s exam, graded on our server, with three attempts at each
  • A year of access to the whole track, charged once rather than renewed
  • A certificate of completion for every course, and one for AI Cyber Security Foundations as a whole

List price $1,099

$999

per candidate, paid once, for all seven

Introductory offer: $100 off until December 31, 2026

Add to CartGet Started

Paid once, for each named candidate. Add it to your cart, on its own or with anything else, and pay by card; your organization is set up when you pay. One price, not banded by volume: this is a credential bought for a person, not a deployment. A proctored sitting, for the verified credential, is priced separately: $499 with AI proctoring or $750 with a human invigilator, bought in the portal once a seat is assigned. A retake is $249. Questions first? Ask us.

How the price compares

OffSec's AI red-teaming course, AI-300, costs $1,749 for 90 days of access with one exam attempt, or $2,749 a year through their Learn One plan with two. Either way that is one course, taught with hands-on labs and examined by a 24-hour proctored practical. AI Cyber Security Foundations lists at $1,099 for a year of all seven courses, with three attempts at each exam, and is $999 with our introductory offer until December 31, 2026. Its courses are written rather than lab-based, and its exams are scenario questions graded on our server.

Prices as published on offsec.com on September 23, 2026. OffSec, AI-300 and Learn One are OffSec's names. FluentOrg is independent and not affiliated with or endorsed by OffSec.

A second course

AI Cyber Security Operations

For the team that watches AI in production and answers the page at two in the morning: security operations analysts, detection engineers, incident responders, threat hunters and the engineers who run the features. Five courses, 8.5 hours, built on the first course of AI Cyber Security Foundations.

  1. Course 1 · 90 minutes

    Telemetry: What an AI System Must Record

    What an AI system has to write down for an incident to be investigable at all, when the cause sits in a retrieved document and the model will not make the same decision twice.

    You leave with a telemetry specification for one feature, and a retention decision agreed with privacy before an incident needs it.

  2. Course 2 · 120 minutes

    Detection Engineering for AI Systems

    Detections for what an attack causes rather than what it says: authority exceeded, retrieval across customers, canaries, action anomalies and egress, each written and tested like code.

    You leave with a detection set for one feature, each rule with its test cases, its severity and the runbook it opens.

  3. Course 3 · 90 minutes

    Threat Hunting in AI Systems

    Hypothesis-led hunts for what has already happened and not yet fired: poisoned content in the index, leaked keys, unapproved AI use, agent drift and quiet cross-customer retrieval.

    You leave with a hunt log, and the detections the hunts became.

  4. Course 4 · 120 minutes

    Incident Response for AI Systems

    Containment by taking authority away rather than arguing with the model, evidence from a system that will not repeat itself, and scoping with the six questions every incident asks.

    You leave with an incident runbook: the containment levers for one feature, who may pull each, and the playbooks that use them.

  5. Course 5 · 90 minutes

    Recovery, Review and the Tabletop

    Removing what an incident left from every copy, proving the fix as a rate, restoring authority in stages, and the tabletop that finds the gaps before an incident does.

    You leave with a post-incident review template and a tabletop your team has already run once.

Its own package

AI Cyber Security Operations

  1. Telemetry: What an AI System Must Record
  2. Detection Engineering for AI Systems
  3. Threat Hunting in AI Systems
  4. Incident Response for AI Systems
  5. Recovery, Review and the Tabletop
  • All five courses, 8.5 hours of study
  • Every course’s exam, graded on our server, with three attempts at each
  • A year of access, charged once rather than renewed
  • A certificate for AI Cyber Security Operations, valid for three years, and a proctored sitting as an option for the verified credential

List price $799

$699

per candidate, paid once, for all five

Introductory offer: $100 off until December 31, 2026

Add to CartGet Started

Paid once, for each named candidate. Add it to your cart, on its own or with anything else, and pay by card; your organization is set up when you pay. On its own, or beside the AI Cyber Security Foundations track by the same candidate. A proctored sitting, for the verified credential, is priced separately: $499 with AI proctoring or $750 with a human invigilator, bought in the portal once a seat is assigned. A retake is $249. Questions first? Ask us.

How the price compares

OffSec sells its security operations course the way it sells every 200-level course: $1,749 for 90 days with one exam attempt, or $2,749 a year through Learn One with two, taught with hands-on labs. Practical DevSecOps’ AI security certification is $1,099 with labs and one attempt, and does not teach detection or incident response. AI Cyber Security Operations is written rather than lab-based, and covers what neither does: detection, hunting and incident response for AI systems.

Prices as published on offsec.com and practical-devsecops.com on September 23, 2026. The names are their owners’. FluentOrg is independent and not affiliated with or endorsed by either.

A certificate for every course, and one for the program.

Finish a course and pass its exam, and its certificate of completion is yours to claim. Finish all seven and you can claim one for AI Cyber Security Foundations as a whole too. Each certificate runs for three years, because what counts as current in AI security moves faster than most subjects, and anyone can check one by its id.

The seven courses are 13 hours of stated study time, yours to self-report as continuing professional education, up to 13 CPE hours. The hours are self-reported: FluentOrg is not a registered CPE sponsor with any certifying body, including the National Association of State Boards of Accountancy (NASBA), and each certifying body decides what it accepts. What you can claim

Check a Certificate

Ready to start?

Add AI Cyber Security Foundations to your cart for each candidate, on its own or with anything else, and pay by card, once.

Add to Cart Ask Us First