Article 15: Accuracy, Robustness and Cybersecurity

Official Statutory Text

Article 15 Accuracy , robustnessand cybersecur ity 1. High-risk AI systems shall be designed and developed in suchaway that they achieve an appropriate level of accuracy , robustness, and cybersecur ity , and that they perform consistently in those respects throughout their lifecy cle. 2. To address the technical aspects of how to measure the appropriate levels of accuracy and robustness setout in paragraph 1 and any other relevant performance metrics, the Commission shall, in cooperation with relevant stakeholdersand organisations suchas metrology and bench marking author ities, encourage , as appropriate, the developmentof benchmarksand measurement methodologies. 3. The levels of accuracy and the relevant accuracy metrics of high-risk AI systems shall be declared in the accompanying instructions of use. 4. High-risk AI systems shall be as resilient as possible regard inger rors, faults or inconsistencies that may occur within the system or the environment in which the system operat es, in particular due to their interaction with natural persons or other systems. Technical and organisational measures shall betaken in this regard . The robustness of high-risk AI systems maybe achieved through technical redundancy solutions, which may include backuporf ail-safe plans. High-risk AI systems that continue to learnafter being placed on the marketor put into service shall be developed in suchawayas to eliminate or reduce as far as possible the risk of possibly biased outputs influencing in put forfuture operations (feedback loops), and as to ensure that any suchfeedback loops are duly addressed withappropriate mitigation measures. 5. High-risk AI systems shall be resilient against attempts by unauthorised third par ties to alter their use, outputs or performance by exploiting system vulnerabilities. The technical solutions aiming to ensure the cybersecur ity of high-risk AI systems shall be appropriate to the relevant circumstancesand the risk s. The technical solutions to address AI specific vulnerabilities shall include, where appropriate, measures to prevent, detect, respond to, resolve and control for attacks trying to manipulate the training data set (data poisoning), or pre-trained components used in training (model poisoning), in puts designed to cause the AI model to make a mistake (adversar ial examples or model evasion), confidentiality attacks or model f laws. SECTION 3 Obligations of providersand deployers of high-risk AI systemsand other par ties