{"slug":"eu-ai-act-gpai-systemic-risk","id":"AIPT-OBL-0019","permanent_url":"https://aipolicytracker.org/id/AIPT-OBL-0019","title":"Manage systemic risk for high-impact general-purpose models","short_title":"GPAI models with systemic risk","category":"safety_testing","summary":"A general-purpose model is presumed to have systemic risk when the cumulative compute used for training exceeds 10^25 floating-point operations, or when the Commission designates it. Providers must notify the Commission, perform model evaluations including adversarial testing, assess and mitigate systemic risks, track and report serious incidents, and ensure adequate cybersecurity for the model and infrastructure.\n","practical_action":"Track training compute against the threshold and prepare a safety and security framework before launch.\n","is_binding":true,"applies_from":"2025-08-02","section":"Articles 51–56","source_reference":"Articles 51, 52 and 55","official_source_url":"https://eur-lex.europa.eu/eli/reg/2024/1689/oj","review_status":"pending_review","confidence_level":"medium","last_verified_at":null,"reviewed_by":null,"actors":["gpai_provider"],"sectors":["cross_sector"],"use_cases":["generative_ai"],"applicability":[{"description":"Providers of general-purpose AI models meeting the systemic-risk criteria.","actors":["gpai_provider"],"conditions":null}],"evidence_examples":[{"title":"Model evaluation and red-teaming reports","description":null,"artifact_type":"report"},{"title":"Commission notification record","description":null,"artifact_type":"record"}],"framework_mappings":[{"framework":"nist_ai_rmf","framework_name":"NIST AI RMF 1.0","reference":"MEASURE 2.x; NIST AI 600-1","note":"Evaluation and red-teaming practices.","confidence_level":"medium","is_original":true}],"controls":[{"slug":"frontier-model-safety-framework","title":"Frontier model safety and security framework","kind":"policy","relationship":"satisfies","note":"Compute tracking, notification and the safety and security framework.","confidence_level":"medium","url":"https://aipolicytracker.org/controls/frontier-model-safety-framework"},{"slug":"genai-red-team-testing","title":"Adversarial and red-team testing for generative AI","kind":"technical","relationship":"satisfies","note":"Model evaluation including adversarial testing.","confidence_level":"medium","url":"https://aipolicytracker.org/controls/genai-red-team-testing"},{"slug":"ai-incident-management-and-reporting","title":"AI incident management and regulatory reporting","kind":"process","relationship":"supports","note":"Tracking and reporting serious incidents to the AI Office.","confidence_level":"medium","url":"https://aipolicytracker.org/controls/ai-incident-management-and-reporting"}],"url":"https://aipolicytracker.org/obligations/eu-ai-act-gpai-systemic-risk","policy":{"slug":"eu-ai-act","title":"Regulation (EU) 2024/1689 — Artificial Intelligence Act","status":"partially_applicable","jurisdiction":"eu","url":"https://aipolicytracker.org/policies/eu-ai-act"}}