OFFICIAL ISC2 AI security exam guidance doc
ISC2 published this yesterday. It maps out exactly how AI security concepts show up across the CISSP exam. This is NOT a new exam outline. The current outline (April 2024) already has AI baked in. But this document spells out the specifics so you know what to expect. The big picture: AI isn't a separate topic. It's woven into everything from risk management (Domain 1) to software development security (Domain 8). A few things that stood out to me: - You need to know about protecting training data and model weights (Domain 2) - Prompt injection and adversarial attacks are fair game (Domain 3) - AI red teaming is now part of security testing (Domain 6) - Managing identities for AI agents and service accounts - least privilege still applies (Domain 5) - Model drift and AI in the SOC are covered in operations (Domain 7) If you're studying right now, don't panic. Most of this maps to concepts you already know -- just applied to AI systems. But you should absolutely be familiar with terms like data poisoning, adversarial attacks, algorithmic bias, model drift, and prompt injection. On our end we're going to keep weaving more AI-focused questions into the https://cissp.app and bringing more of this into our study group discussions. I attached the PDF if you want to read the full thing.