Glossary term
Glossary term
Safety and Alignment
A general-purpose AI model with high-impact capabilities or risk characteristics that may create significant effects across markets, sectors, or society and therefore requires stronger governance. Governance emphasis shifts toward capability evaluation, serious incident reporting, cybersecurity, adversarial testing, and downstream risk communication.
EU AI Act Article 51 designates a GPAI model as having systemic risk when training compute exceeds 10^25 floating point operations.
Under Article 55, providers of GPAI with systemic risk must perform state-of-the-art model evaluations including adversarial testing and report serious incidents to the AI Office.
OpenAI's GPT-4 class models, Anthropic's Claude Opus, and Google DeepMind's Gemini Ultra are widely considered to meet systemic-risk thresholds under EU AI Act criteria.