EU_AI_ACT_ART15_ACCURACY_ROBUSTNESS_001

rules/regulations/eu/eu_ai_act_article15_accuracy_robustness.yaml

Citation

Regulation (EU) 2024/1689, Article 15(1): High-risk AI systems shall be designed and developed in such a way that they achieve an appropriate level of accuracy, robustness, and cybersecurity, and that they perform consistently in those respects throughout their lifecycle. Article 15(3): The levels of accuracy and the relevant accuracy metrics of high-risk AI systems shall be declared in the accompanying instructions of use. Article 15(4), first subparagraph: High-risk AI systems shall be as resilient as possible regarding errors, faults or inconsistencies that may occur within the system or the environment in which the system operates, in particular due to their interaction with natural persons or other systems. Technical and organisational measures shall be taken in this regard. Article 15(5): High-risk AI systems shall be resilient against attempts by unauthorised third parties to alter their use, outputs or performance by exploiting system vulnerabilities. The technical solutions aiming to ensure the cybersecurity of high-risk AI systems shall be appropriate to the relevant circumstances and the risks. 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), inputs designed to cause the AI model to make a mistake (adversarial examples or model evasion), confidentiality attacks or model flaws.

Condition

Requires high-risk AI systems to achieve appropriate levels of accuracy, robustness, and cybersecurity throughout their lifecycle, with resilience against errors, faults, adversarial manipulation, and model poisoning. Article 15.

Layer 1 detects

"without accuracy", "without robustness" or "without cybersecurity" followed by "testing", "validation" or "assessment"; "no adversarial", "no robustness" or "no accuracy" followed by "testing", "assessment" or "evaluation"; "vulnerable to adversarial", "vulnerable to model poisoning", "vulnerable to manipulation" or "vulnerable to attack"; "no cybersecurity" or "no security" followed by "protection", "measure" or "assessment"; "accuracy not declared", "not documented", "not tested" or "not validated"; "adversarial attack" or "adversarial manipulation" followed by "not tested", "not assessed" or "not mitigated"; "model poisoning not addressed", "not mitigated" or "not tested"; "deploy" or "deploying" followed by "without accuracy", "without robustness" or "without security", then "testing" or "validation"; and "no resilience against" or "no resilience to" followed by "errors", "faults" or "attacks". Case is ignored, the words must follow each other with no other word between them unless stated, and a phrase can start inside a longer word. It matches wording, not meaning: it misses the practice told in other words ("Remove the fallback path so the triage model output goes straight to the queue even when it crashes."), it does not check that the system is high-risk, and it fires on text that only names the practice, such as a refusal. Whether a text does the practice is judged in Layer 2.

Condition type

regex; semantic condition in the file, not run by the engine

Status in the file

approved