CyCon 2026 Series – New Technologies, Armed Conflict, and International Humanitarian Law

Articles of War | Maria Tolppa

NATO's 18th International Conference on Cyber Conflict in Tallinn highlighted how integrated artificial intelligence and cyber capabilities are transforming modern command structures and international humanitarian law compliance. Operational tools such as the Lavender and Gospel targeting systems demonstrate that algorithmic processing drastically accelerates target selection but introduces severe risks when underlying intelligence feeds are corrupted or unverified.

Integrating these digital capabilities across multi-domain operational environments creates severe structural strain on established legal frameworks like Article 36 of Additional Protocol I. Legal compliance demands continuous lifecycle governance. Scholars including Dr. Anke Allenhöfer, Colonel Inna Zavorotko, and Professor Rain Liivoja emphasize that human-in-the-loop oversight requires genuine operational time, complete contextual understanding, and direct command authority to override automated recommendations. Because formal international treaty reform remains unlikely, regulatory compliance depends on practical safeguards embedded within procurement contracts, continuous operational training, updated military manuals, and public-private technical partnerships.

Comment

The reliance on data-driven targeting algorithms during active operations strains the structural mechanics of Article 36 of Additional Protocol I. Pre-deployment reviews traditionally evaluated fixed technical parameters of physical weaponry rather than dynamic software that evolves via iterative machine learning. When operational tools like Lavender process probabilistic indicators, legal compliance shifts from a static clearance point to a continuous verification process. This structural friction reveals that conventional weapon review mechanisms are poorly configured for software systems that mutate post-fielding.

Translating Article 36 obligations into real-time operational environments requires audit logging and algorithmic rollback protocols directly within military decision-support architectures. Without verifiable technical documentation of data provenance, commanders operating under intense time pressure face automation bias when evaluating probability outputs generated by systems like Gospel. Consequently, software-defined platforms evaluated under Article 36 require embedded logging mechanisms to preserve decision auditability during joint operations.

Strategic Question for Discussion

If machine-learning algorithms like Gospel continuously update their target identification logic in active combat, which framework better preserves legal auditability — embedding automated system-level logging, or enforcing mandatory operational pauses for human re-evaluation?

The operational trajectory suggests that enforcing mandatory human re-evaluation creates severe friction in high-tempo multi-domain combat environments. My assessment is that military command structures will increasingly favour embedded technical audit logging to maintain Article 36 compliance without sacrificing targeting velocity. This approach shifts legal verification directly into software development pipelines, making real-time data provenance the core foundation of algorithmic accountability.

Share your assessment in the comments below.

To read in details click here