EU-AIA-15-09
Requirement
15 — Accuracy, robustness and cybersecurity
Address AI-specific vulnerabilities with technical solutions
Description
Technical solutions to address AI-specific vulnerabilities must include, where appropriate, measures to prevent, detect, respond to, resolve and control for attacks trying to manipulate training datasets (data poisoning), pre-trained components (model poisoning), adversarial examples, confidentiality attacks or model flaws.
Full Analysis & Evidence Requirements
Sign in to view the full obligation text, AI-generated applicability analysis, evidence checklists, and compliance mapping.
Sign In to ViewRelated Obligations
EU-AIA-15-01
Requirement
Design and develop for appropriate accuracy, robustness and cybersecurity
EU-AIA-15-02
Transparency
Declare accuracy levels and metrics in instructions
EU-AIA-15-03
Requirement
Ensure resilience to errors, faults and inconsistencies
EU-AIA-15-04
Requirement
Implement technical and organisational resilience measures
EU-AIA-15-05
Requirement
Eliminate or reduce biased feedback loops in learning systems
EU-AIA-15-06
Requirement
Address feedback loops with appropriate mitigation measures
EU-AIA-15-07
Requirement
Ensure resilience against unauthorized third-party alteration
EU-AIA-15-08
Requirement
Implement appropriate cybersecurity technical solutions
Map this obligation to your AI systems
ReguLume automatically maps regulatory obligations to your system inventory, identifies compliance gaps, and generates remediation plans.
Get Started