From emerging regulation to deal mechanics, we write about the questions founders and investors actually ask — practical analysis you can put to work.
The EU AI Act is the world’s first horizontal AI law: a risk-based regulation applying in tiers between 2025 and 2028, with extraterritorial reach over any provider whose AI system or output is used in the EU — including Turkey-based companies.
AI safety is the field of research, engineering, and policy aimed at preventing AI systems from causing harm—whether through misuse, accident, or misalignment of objectives.
The Transformer is the neural network architecture underlying virtually all modern LLMs (GPT, Claude, Gemini, Llama), image generation models, code models, and other foundation models.
AI alignment is the field of research and engineering dedicated to ensuring that AI systems pursue the goals their designers and users actually want, rather than misinterpreting instructions, exploiting reward functions, or producing harmful outputs.
Hallucination in AI refers to the phenomenon where a model—typically an LLM or generative AI—produces output that appears plausible but is factually incorrect, nonsensical, or unsupported by provided context.
Agentic AI is the design paradigm in which AI systems act autonomously toward goals — planning, using tools and executing multi-step workflows. See our canonical entry “AI agent (agentic AI)” for the full legal analysis.
An AI agent is an AI system that autonomously plans and executes multi-step tasks toward a goal — calling tools, browsing, transacting — raising attribution, oversight and liability questions that ordinary chatbots do not.
Fine-tuning adapts a pre-trained model with your own data. Where the legal lines run: when you become the AI Act provider, what happens to personal data in weights, and what your model licence actually permits.
Prompt engineering is the practice of designing inputs to LLMs and other generative AI systems to elicit accurate, useful, and consistent outputs. It combines empirical experimentation with model behavior, structured input design, and increasingly automated optimization.
An embedding is a dense vector representation of data—text, images, audio, code—that captures semantic meaning in a continuous numerical space.
