ISC2 publishes the outline of its new AI security certification exam

The exam covers AI security from design to governance

ISC2 says the certification is for people who secure the design, deployment and maintenance of AI and machine learning systems.

The exam's areas run from AI concepts and security fundamentals to governance, risk and compliance. They also cover data and model security, secure architecture, operations and incident response, and AI threats and testing. The list below gives each area's full name.

The outline draws on input from nearly 10,000 volunteer professionals worldwide. Casey Marks, ISC2's chief operating officer, said the pilot gives critical evidence on whether the exam measures AI security competence fairly and reliably.

Four years of experience, or an existing ISC2 credential

Candidates need at least four years of paid full-time experience in cyber security or AI-related fields. Two of those years must be in cyber security, and one year must be in two or more of the exam's areas.

An active CISSP, ISSAP, ISSEP, ISSMP, CCSP or CSSLP credential can replace the whole experience requirement. Candidates do not need to hold an ISC2 certification.

The pilot exam runs from mid-December

Approximately 3,000 qualified volunteers will be selected to sit the pilot exam from mid-December to late January. Anyone who wants to take part can submit their interest through ISC2's volunteer form by 4 November.

If the pilot is successful, ISC2 says the certification will become operational and launch in 2027.

In the UK, AI security training is in a voluntary code

The UK government's Code of Practice for the Cyber Security of AI is voluntary. It is meant to help create a global standard at the European Telecommunication Standards Institute (ETSI).

Its first principle is to raise awareness of AI security threats and risks. The code says cyber security training shall include AI security content, tailored to the roles and responsibilities of staff.