OffSec is the company behind the OSCP, still the best known practical penetration testing certification. AI-300 is its move into AI security, leading to the OSAI credential. It became available through Learn One subscriptions on 31 March 2026. OffSec pitches it as the natural pivot for OSCP holders: people who already know how to break into systems and now need to test LLM applications, agents and the infrastructure around models. The pricing is the first thing to understand.
A course and certification bundle starts at $1,749. Learn One costs $2,749 for a year and lets you choose one 200-level or 300-level course, with two attempts at that exam plus attempts at two smaller certifications. If you want OSAI, Learn One is the better value, but either way it is a lot of money. I have not been able to judge the exam itself, and I would be cautious of anyone who claims to after a few months. OffSec exams are known for being hard and practical, and if AI-300 follows that pattern it will mean something.
What I can say is that the topics overlap heavily with the Hack The Box AI Red Teamer path, which costs a fraction of this and was built with Google. TryHackMe's AI path is cheaper again. So the question is what you are paying for. With OffSec you mostly pay for the name. For a consultant whose clients ask for OffSec credentials, or an employee whose training budget is use it or lose it, that is a legitimate reason.
For someone paying from their own pocket, it is hard to justify as the first AI security course. My take: probably solid, certainly expensive, and too new to recommend with confidence. Let others test it for a year unless someone else is paying.