OWASP AIBOM co-leads release RAID-AI, a free ransomware readiness framework for AI systems
Behnaz Karimi and Yuvaraj Govindarajulu, co-leads of the OWASP AI Bill of Materials (AIBOM) initiative, have released RAID-AI (Ransomware Assessment and Incident Defense Framework for AI Systems). It is a free web console that helps security teams self-assess their resilience to ransomware attacks on AI models, data and supporting infrastructure. The authors developed it over two years and had planned to present it at the ISC2 Security Congress in Denver (26-28 October), but visa issues kept Karimi from attending. Karimi cited an April 2025 report on ransomware actors using data poisoning, and an alleged June 2026 attack on Novo Nordisk in which hackers claimed to have stolen AI models, as signs that such attacks are emerging. This is a defensive tooling release, not a new threat or incident.
Detalhes técnicos
- The assessment is manual. Users rate controls on a 0-5 maturity scale, with levels such as basic, advanced and high. The tool does not connect to or scan systems and does not verify security posture independently.
- Controls are split into general ones (recovery, training, sensitive-data protection) and AI-specific ones (data, tools, runtime).
- Controls map to NIST CSF 2.0 and CISA CPG 2.0. Adversarial attacks can be mapped through MITRE ATT&CK, MITRE ATLAS and NIST Adversarial.
- The console provides dashboards and downloadable PDF reports.
Remediação
- Run the assessment in workshops that bring together security managers, engineers and AI-security specialists, and use the PDF report to prioritize gaps.
- The authors plan to add implementation videos and blueprints, and a similar tool for ransomware threats against agentic AI systems.
Linha do tempo
-
OWASP AIBOM co-leads release RAID-AI, a free ransomware readiness framework for AI systems
ler a análise desta fonte ocultar
Behnaz Karimi and Yuvaraj Govindarajulu, co-leads of the OWASP AI Bill of Materials (AIBOM) initiative, have released RAID-AI (Ransomware Assessment and Incident Defense Framework for AI Systems). It is a free web console that helps security teams self-assess their resilience to ransomware attacks on AI models, data and supporting infrastructure. The authors developed it over two years and had planned to present it at the ISC2 Security Congress in Denver (26-28 October), but visa issues kept Karimi from attending. Karimi cited an April 2025 report on ransomware actors using data poisoning, and an alleged June 2026 attack on Novo Nordisk in which hackers claimed to have stolen AI models, as signs that such attacks are emerging. This is a defensive tooling release, not a new threat or incident.
Technical Details
- The assessment is manual. Users rate controls on a 0-5 maturity scale, with levels such as basic, advanced and high. The tool does not connect to or scan systems and does not verify security posture independently.
- Controls are split into general ones (recovery, training, sensitive-data protection) and AI-specific ones (data, tools, runtime).
- Controls map to NIST CSF 2.0 and CISA CPG 2.0. Adversarial attacks can be mapped through MITRE ATT&CK, MITRE ATLAS and NIST Adversarial.
- The console provides dashboards and downloadable PDF reports.
Remediation
- Run the assessment in workshops that bring together security managers, engineers and AI-security specialists, and use the PDF report to prioritize gaps.
- The authors plan to add implementation videos and blueprints, and a similar tool for ransomware threats against agentic AI systems.