arXiv: Landseer: Exploring the Machine Learning Defense Landscape

This publication, titled Landseer: Exploring the Machine Learning Defense Landscape, is a technical research paper from arXiv that maps current adversarial attack and defense methods for machine learning systems. It does not represent a new regulation or binding legal change, but it provides a structured taxonomy of vulnerabilities and countermeasures relevant to AI safety frameworks, particularly under the EU AI Act’s requirements for robustness and risk management.

The primary audience is organizations deploying high-risk AI systems, especially in sectors like finance, healthcare, critical infrastructure, and autonomous systems. Compliance teams in these sectors should review the paper’s defense categories to assess whether their current AI models are protected against known evasion, poisoning, and extraction attacks. The paper also highlights gaps in standardized testing, which may inform future conformity assessments.

Compliance teams should treat this as a technical reference for updating their AI risk management procedures. Next steps include mapping the paper’s defense strategies to their own model inventories, evaluating whether existing testing protocols cover the identified attack vectors, and documenting any gaps for internal audit or regulatory reporting. Engaging with the technical teams to implement or test recommended defenses is advisable, especially for systems subject to third-party conformity assessments under the AI Act.