CAISP Certification Guide 2026: AI Security vs AI Governance

Quick answer

The Certified AI Security Professional (CAISP) from Practical DevSecOps is a hands-on AI security credential. Its exam is a 6-hour practical challenge with 5 real-world tasks, followed by a written report within 24 hours. It costs about $1,099, including 60 days of lab access. It covers LLM attacks (MITRE ATLAS, OWASP LLM Top 10), AI supply-chain security, threat modelling and an overview of governance frameworks. It suits security engineers. AI security credentials complement governance credentials such as the AIGP. They don't replace them.

An earlier version of this article claimed governance certifications were "dying". That isn't supported by evidence, and we've corrected it. Demand for both AI security and AI governance skills is growing, and they serve different roles. Here's how the CAISP fits.

CAISP at a glance

Item	Detail
Provider	Practical DevSecOps
Format	Hands-on: 6-hour practical exam (5 tasks) + report within 24 hours; no multiple choice
Price	About $1,099, including 60 days of browser-based labs and one exam attempt
Topics	LLM attacks mapped to MITRE ATLAS; OWASP Top 10 for LLMs (prompt injection, training data poisoning, etc.); AI supply-chain security; STRIDE-based AI threat modelling; AI in DevOps pipelines; overview of NIST AI RMF, ISO/IEC 42001 and the EU AI Act
Best for	Application security, DevSecOps and red-team practitioners securing AI systems

Details are as published by the provider. Check its site for current pricing and syllabus.

AI security vs AI governance credentials

	CAISP	CompTIA SecAI+	AIGP
Focus	Hands-on AI security	AI security fundamentals	AI governance (law, frameworks, lifecycle)
Format	Practical labs exam	60 questions, 60 minutes (multiple choice + performance-based)	100 multiple-choice questions, 2 h 45 min
Launched	Established provider credential	17 February 2026 (CY0-001)	2024 (BoK v2.1 from Feb 2026)
Typical role	Security engineer, AI red teamer	Security analyst	AI governance lead, privacy/legal/risk professional

Who should choose which

You secure AI systems hands-on (testing, hardening, threat modelling): CAISP or a similar practical credential.
You're a security analyst new to AI: CompTIA SecAI+ as a foundation.
You govern AI (policy, risk, compliance, legal, programme): AIGP.
You lead AI security and governance together: a combination, such as AIGP plus a practical security credential.

Why both matter: security finds and fixes vulnerabilities such as prompt injection. Governance decides acceptable risk, who's accountable and what the law requires, and makes sure security findings change release decisions. Organisations need both. See AI red teaming for governance teams.

What hands-on AI security work involves

Threat modelling AI systems: mapping where prompts, retrieved content, tools and outputs cross trust boundaries.
Testing for prompt injection (direct and indirect) and data leakage in LLM applications.
Securing the AI supply chain: verifying model and dataset sources, scanning dependencies, controlling model registries.
Hardening agents: least-privilege tool permissions, output validation before actions, sandboxing.
Integrating AI checks into CI/CD, so tests run on every model or prompt change.

The OWASP Top 10 for LLM applications (2025)

AI security credentials such as CAISP expect you to know the main LLM application risks. The 2025 OWASP list is a good map:

#	Risk	In plain terms
LLM01	Prompt injection	Inputs that override the system's instructions, directly or through documents and web pages
LLM02	Sensitive information disclosure	The model reveals personal data, secrets or confidential business information
LLM03	Supply chain	Compromised models, datasets, plugins or libraries
LLM04	Data and model poisoning	Tampered training, fine-tuning or embedding data
LLM05	Improper output handling	Passing model output to other systems without validation (for example, into code or SQL)
LLM06	Excessive agency	Giving an AI agent more permissions or autonomy than it needs
LLM07	System prompt leakage	Exposing instructions or secrets placed in the system prompt
LLM08	Vector and embedding weaknesses	Weaknesses in retrieval systems, such as access control gaps in a vector store
LLM09	Misinformation	Confident but false outputs that people rely on
LLM10	Unbounded consumption	Uncontrolled resource use, leading to cost spikes or denial of service

MITRE ATLAS in brief

MITRE ATLAS is a knowledge base of adversary tactics and techniques against AI systems, modelled on the well-known MITRE ATT&CK framework. Security teams use it to threat model AI systems, plan red team exercises and map detections. Expect to know what it is and how it's used rather than every technique.

A study plan for the CAISP

Weeks 1–2: AI and LLM basics (training, inference, embeddings, RAG, agents) if you're coming from general security.
Weeks 3–4: work through the OWASP LLM Top 10 and MITRE ATLAS, with lab exercises on prompt injection and output handling.
Week 5: supply chain security for models and datasets; threat modelling an AI application end to end.
Week 6: practise writing clear findings reports, since the exam includes a written report.

CAISP vs CompTIA SecAI+

	CAISP	CompTIA SecAI+
Format	Hands-on practical exam plus report	60 questions in 60 minutes
Pass mark	Practical tasks and report assessment	600 on a 900-point scale
Best for	Practitioners who want to show hands-on skill	A vendor-neutral, exam-based credential

How security and governance teams work together

Activity	Security (e.g. CAISP holder)	Governance (e.g. AIGP holder)
Red teaming	Designs and runs attacks; reports findings	Sets scope by risk; makes sure findings affect release decisions
Vendor AI	Assesses technical security of the model or API	Contract terms, data use, regulatory classification
Incidents	Containment and technical root cause	Notification duties, accountability, process changes
Agents	Implements permission limits and sandboxing	Defines which actions need human approval

See AI red teaming for governance teams and the best AI governance certifications.

Adding governance to your security skills?

The AIGP Complete Pack covers the EU AI Act, NIST AI RMF, ISO/IEC 42001 and lifecycle controls in 300 BoK v2.1 questions.

Get the Complete Pack

Frequently asked questions

What is the CAISP certification?

A hands-on AI security credential from Practical DevSecOps, assessed through a 6-hour practical exam and a written report.

How much does the CAISP cost?

About $1,099, including 60 days of lab access and one exam attempt, according to the provider.

Is the CAISP better than the AIGP?

They serve different roles. The CAISP is for securing AI systems. The AIGP is for governing them.

Are AI governance certifications losing value?

There's no evidence of that. IAPP's 2025–26 data links AI governance responsibility with higher pay, and regulation is increasing demand.

What is the OWASP Top 10 for LLM applications?

A list of the ten most important security risks for applications built on large language models, from prompt injection and sensitive information disclosure to excessive agency and unbounded consumption. The 2025 edition is the current version.

What is MITRE ATLAS?

A knowledge base of adversary tactics and techniques against AI systems, modelled on MITRE ATT&CK. Security teams use it for threat modelling and red team planning.

Sources

Last reviewed: October 2026.