←All the glossaryGlossary

AI

Artificial intelligence both as a risk object and as a working tool: architectures, model-specific failures, data governance, and the European framework that applies to them.

60 terms

Adversarial examples Agency theory Agentic AI AI model supply chain AI regulatory sandbox Algorithmic accountability Algorithmic bias Algorithmic black box Algorithmic discrimination Algorithmic hallucination Algorithmic labor (Lalgo)Algorithmic monoculture (software single point of failure)ALM model (Autor-Levy-Murnane)Automation bias Autonomous agent Benchmark contamination CAPEX / OPEX Carbon footprint of model training Cognitive marginal cost Context window Data governance Data leakage through assistant prompts Data poisoning Differential privacy Digital twin Embodied AI Emergent abilities Explainability (explainable AI)Federated learning Fine-tuning Flash Crash Foundation model Generated content watermarking Guardrails Human in the loop Human oversight ISO/IEC 42001 Jailbreak Large Language Model (LLM)Liability for AI hallucinations Machine unlearning Membership inference attack Model card Model distillation Model drift Model extraction attack Model inversion attack Multi-agent systems NIST AI Risk Management Framework Oracle (parametric insurance)Polanyi's paradox Proxy discrimination ReAct (Reasoning and Acting)Reinforcement learning from human feedback Retrieval-augmented generation (RAG)Synthetic data Systemic risk of a general-purpose AI model Technological hold-up Text and data mining for training data Zero-Employee Firm

Other categories

Cyber Insurance Reinsurance Actuarial science Law & regulation Finance Markets Special risks Sustainable insurance Infrastructure & data centers Space risks Alternative risk transfer Transport Life & protection Claims & settlement