{"slug":"eu-ai-act-art-55-systemic-risk-cybersecurity","id":"AIPT-OBL-0046","permanent_url":"https://aipolicytracker.org/id/AIPT-OBL-0046","title":"Providers of systemic-risk GPAI models must secure the model and its infrastructure","short_title":"Systemic-risk model cybersecurity","category":"accuracy_robustness_security","summary":"Providers of general-purpose AI models with systemic risk must ensure an adequate level of cybersecurity protection for the model and for the physical infrastructure it runs on, which in practice covers protection of model weights, training and inference environments, and access controls against theft, tampering and unauthorised release. Codes of practice and harmonised standards may be used to show compliance.\n","practical_action":"Apply a documented security standard to weight storage, training clusters and inference endpoints, with penetration tests and insider-risk controls.\n","is_binding":true,"applies_from":"2025-08-02","section":"Articles 51–56","source_reference":"Article 55(1)(d)","official_source_url":"https://eur-lex.europa.eu/eli/reg/2024/1689/oj","review_status":"verified","confidence_level":"high","last_verified_at":"2026-09-26","reviewed_by":null,"actors":["gpai_provider"],"sectors":["cross_sector"],"use_cases":["generative_ai"],"applicability":[{"description":"Providers of general-purpose AI models classified as having systemic risk under Article 51.","actors":["gpai_provider"],"conditions":null}],"evidence_examples":[{"title":"Model and infrastructure security assessment","description":null,"artifact_type":"report"},{"title":"Weight access control records","description":null,"artifact_type":"record"}],"framework_mappings":[{"framework":"iso_27001","framework_name":"ISO/IEC 27001:2022","reference":"Clause 8.1; Annex A 5.15, A 8.24","note":"Access control and cryptography applied to model assets.","confidence_level":"medium","is_original":true},{"framework":"nist_ai_rmf","framework_name":"NIST AI RMF 1.0","reference":"MEASURE 2.7, MANAGE 2.2","note":"Security and resilience of the system.","confidence_level":"medium","is_original":true},{"framework":"mitre_atlas","framework_name":"MITRE ATLAS","reference":"AML.M0005, AML.M0001","note":"Control access to models and limit artifact release.","confidence_level":"medium","is_original":true}],"controls":[{"slug":"accuracy-robustness-security-testing","title":"Accuracy, robustness, fairness and security testing","kind":"technical","relationship":"satisfies","note":"Security testing of the model and its serving infrastructure.","confidence_level":"medium","url":"https://aipolicytracker.org/controls/accuracy-robustness-security-testing"},{"slug":"frontier-model-safety-framework","title":"Frontier model safety and security framework","kind":"policy","relationship":"supports","note":"Framework commits to the weight-security measures.","confidence_level":"high","url":"https://aipolicytracker.org/controls/frontier-model-safety-framework"}],"url":"https://aipolicytracker.org/obligations/eu-ai-act-art-55-systemic-risk-cybersecurity","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"}}