Navigating the professional landscape of AI security requires fluency not just in core technical concepts, but also in the shorthand and jargon used in meetings, reports, and internal communications. Understanding these abbreviations is essential for effective collaboration with development, operations, and business teams. This glossary covers common acronyms and terms you will encounter daily.
Common Abbreviations and Jargon
The following table provides the abbreviation or jargon, its full name or meaning, and its common Hungarian equivalent. This is a living list; the industry evolves, and so does its language.
| Abbreviation / Jargon | Full Term / Meaning | Hungarian Equivalent |
|---|---|---|
| AGI | Artificial General Intelligence. A hypothetical type of AI that can understand, learn, and apply its intelligence to solve any problem, much like a human being. | Általános Mesterséges Intelligencia |
| API | Application Programming Interface. A set of rules and tools for building software and applications; the primary way you will interact with and test most production models. | Alkalmazásprogramozási felület |
| BI | Business Intelligence. The use of data, technology, and analytics to help business leaders make more informed decisions. AI is often a component of modern BI systems. | Üzleti intelligencia |
| DevSecOps | Development, Security, and Operations. An approach that integrates security practices within the DevOps process, aiming to automate security from the start. | DevSecOps (nincs elterjedt magyar megfelelője) |
| E2E | End-to-End. Refers to a process or system from its beginning to its end, such as “end-to-end testing” which validates an entire application workflow. | Végponttól végpontig |
| FP / FN | False Positive / False Negative. Core concepts in model evaluation. An FP is an incorrect positive prediction (e.g., classifying a safe email as spam). An FN is a missed positive (e.g., classifying a malicious file as safe). | Hamis pozitív / Hamis negatív |
| GPU / TPU | Graphics Processing Unit / Tensor Processing Unit. Specialized hardware accelerators crucial for training and running large-scale AI models efficiently. | Grafikus feldolgozóegység / Tenzor feldolgozóegység |
| MLOps | Machine Learning Operations. A set of practices that aims to deploy and maintain machine learning models in production reliably and efficiently. | MLOps (Gépi Tanulási Műveletek) |
| PoC | Proof of Concept. A small-scale project or experiment to test the feasibility of a concept or theory before committing to a full-scale project. | Megvalósíthatósági igazolás / Koncepcióbizonyítás |
| SaaS / PaaS | Software as a Service / Platform as a Service. Cloud computing models. Many AI tools and models are delivered as SaaS (e.g., OpenAI API) or built on PaaS (e.g., AWS SageMaker). | Szoftver mint szolgáltatás / Platform mint szolgáltatás |
| SOTA | State-of-the-Art. The highest level of development or achievement in a particular field at a given time. Used to describe the most advanced models or techniques. | A technika jelenlegi állása / Csúcstechnológia |
| TTP | Tactics, Techniques, and Procedures. A concept from cybersecurity describing the behavior of an adversary. In AI red teaming, this refers to the methods used to attack AI systems. | Taktikák, Technikák és Eljárások |
| UI / UX | User Interface / User Experience. UI is the visual layout of an application, while UX is the overall experience of a person using it. Both are relevant when assessing AI-powered applications. | Felhasználói felület / Felhasználói élmény |
| YoY / QoQ | Year-over-Year / Quarter-over-Quarter. Business terms for comparing performance in a given period against the same period in the previous year or quarter. Often used to measure the impact of AI initiatives. | Éves / Negyedéves összehasonlítás |
Familiarity with this vocabulary is not just about appearing knowledgeable; it’s about precision. When you report a vulnerability tied to an API rate limit issue in a SOTA model during a PoC phase, using the correct jargon ensures your message is understood quickly and accurately by all stakeholders, from engineers to product managers.