A large language model is a neural network trained on very large amounts of text to predict the next word, which lets it answer questions, summarise, translate and write code. LLMs are the most common type of generative AI and of general-purpose AI model under the AI Act.
Generative AIGenerative AI is a family of AI models that produce new content (text, code, images, audio) in response to a prompt, by learning patterns from very large datasets. In finance it is used for drafting, summarising documents, coding help and customer service assistants.
Retrieval-augmented generation (RAG)Retrieval-augmented generation is a technique in which a large language model first searches a trusted document set, such as internal policies, and then answers using the passages it found, ideally citing them. It reduces made-up answers and keeps content current without retraining.
AI hallucinationAn AI hallucination is output from a generative model that sounds plausible but is false or unsupported, such as an invented regulation article or a wrong figure. Controls include grounding answers in sources, human review and testing against known answers.
Generative AIGenerative AI is a family of AI models that produce new content (text, code, images, audio) in response to a prompt, by learning patterns from very large datasets. In finance it is used for drafting, summarising documents, coding help and customer service assistants.
General-purpose AI modelUnder the AI Act, a general-purpose AI model is one trained on large data that can perform a wide range of distinct tasks and be built into many systems, as large language models are. Providers must supply technical documentation and a training data summary; models with systemic risk face extra duties.
AI Act (Artificial Intelligence Act)The AI Act, Regulation (EU) 2024/1689, is the EU law that sets rules for AI systems according to their risk: some practices are banned, high-risk systems face strict requirements, and general-purpose AI models have transparency duties. Credit scoring of individuals is a high-risk use.