AI in Cybersecurity
Working Group

To empower libraries, archives, and museums to harness AI as a tool for strengthening cybersecurity defenses while governing its risks responsibly — ensuring that cultural heritage institutions can protect their digital collections, infrastructure, and communities in an era of AI-powered threats.

• Threat Landscape Mapping — Survey and document cybersecurity threats specific to LAM institutions (ransomware, adversarial attacks on AI classifiers, data poisoning, deepfake content injection) and produce a sector-specific risk register.
• AI-Powered Defense Guidelines — Develop practical guidance on using AI for anomaly detection, access monitoring, and incident response, adapted for institutions with varying levels of technical capacity.
• Content Authenticity & Provenance — Collaborate with the C2PA for G+LAM community to advance standards for verifying the authenticity of digitized and AI-processed heritage content.
• Policy & Procurement Frameworks — Draft model policies and vendor assessment checklists to help institutions evaluate the cybersecurity posture of AI tools and services before adoption.
• Capacity Building — Produce training materials and run workshops to raise cybersecurity awareness for LAM staff at all levels, with focus on AI-specific risks (prompt injection, model misuse, synthetic media).
• Cross-Sector Knowledge Exchange — Facilitate regular knowledge-sharing sessions between LAM practitioners, national cybersecurity agencies, and GLAM technology partners to track the evolving threat environment.

• LAM Cybersecurity Risk Register (Year 1) — an open, living document cataloguing AI-related cybersecurity threats specific to LAM institutions, with severity ratings and mitigation notes.
• Best Practice Guide for AI-Assisted Defense (Year 1–2) — a practical guide covering anomaly detection, access monitoring, and incident response using AI tools, calibrated for small, medium, and large LAM institutions.
• AI Vendor Cybersecurity Assessment Checklist (Year 1) — a ready-to-use procurement tool for evaluating AI product vendors, shared openly under a Creative Commons licence.
• Training Curriculum and Workshop Series (Year 2) — modular training materials for LAM staff covering AI-specific threats; delivered as webinars and hands-on workshops at Fantastic Futures.
• Content Authenticity Position Paper (Year 2) — a collaborative white paper on AI, content provenance, and digital preservation in LAM contexts, contributing to international standards discussions (C2PA, IFLA).

• Use Case 1: Detecting Unauthorized Access to Digital Collections. A national archive uses AI-based anomaly detection (e.g., Darktrace, Vectra AI) to flag unusual access patterns to digitized manuscript collections, triggering automated alerts before data exfiltration occurs.
• Use Case 2: Verifying Authenticity of AI-Processed Scans. A library uses C2PA-compliant tools to attach provenance metadata to digitized images processed by AI enhancement pipelines, enabling downstream users to verify the document’s origin and any AI transformations applied.
• Use Case 3: Evaluating an AI Transcription Vendor. A museum procurement team uses the WG’s vendor checklist to assess a commercial AI transcription provider, identifying data residency risks and negotiating appropriate data processing agreements before deployment.

• NIST AI Risk Management Framework (AI RMF)
• C2PA (Coalition for Content Provenance and Authenticity) standard
• ENISA AI Threat Landscape report
• MITRE ATLAS (Adversarial Threat Landscape for AI Systems)
• OpenCTI for threat intelligence sharing

• NIS2 Directive 2022/2555
• Critical Entities Resilience Directive – UE 2022/2557
• NATO Cyber Defence
• NATO Cognitive Warfair

• Murray, K. & Sternfeld, J. (2026). Content Authenticity and Provenance in the Age of Artificial Intelligence: A Call-to-Action for the LAMs Community. Library of Congress / C2PA for G+LAM.
• ENISA (2023). AI Cybersecurity Challenges: Threat Landscape for AI. European Union Agency for Cybersecurity.
• NIST (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). National Institute of Standards and Technology.
• Cloud Security Alliance (2026). The State of AI Cybersecurity 2026. CSA Research.
• IFLA Statement on Libraries and Artificial Intelligence (latest edition). International Federation of Library Associations and Institutions.
• MITRE ATLAS Knowledge Base — Adversarial Threat Landscape for AI Systems. Practical case studies and mitigations for AI-specific attacks: https://atlas.mitre.org

• C2PA Standard (Content Authenticity)
• NIST AI Risk Management Framework
• ENISA AI Threat Landscape
• MITRE ATLAS (AI adversarial threats)
• Content Authenticity & Provenance White Paper (Library of Congress, 2026)
• IEEE SMC TC Cyber Humanities