← Back to Human Flag

Lawful Operational Safeguards in AI Systems

Surrender Recognition, Compliance Architecture, and International Humanitarian Law

Giovanni Nardacci — Founder, Human Flag Association
UNGM Partner No. 6126 · humanflag.org
Working Paper — May 2026 (Second Revised Edition, June 2026)


Abstract

Contemporary discussions concerning AI safeguards frequently frame such mechanisms as discretionary policy choices, corporate alignment preferences, or operational limitations imposed for safety or reputational reasons. This paper proposes a narrower and more legally grounded distinction: certain categories of safeguards constitute operational mechanisms relevant to compliance with already-existing obligations under International Humanitarian Law ("IHL").

International Humanitarian Law does not merely protect persons hors de combat; it imposes duties of recognition and verification in terms. Article 41(1) of Additional Protocol I protects a person "who is recognized or who, in the circumstances, should be recognized" to be hors de combat; Article 57(2)(a)(i) requires those who decide upon an attack to "do everything feasible to verify" the status of the objective; and Article 36 obliges States to determine, at the stage of study, development, acquisition or adoption of any new weapon, means or method of warfare, whether its employment would be prohibited by existing law. This paper argues that, read together, these provisions make surrender-recognition capability — the technical capacity of a system to identify, preserve, or escalate human surrender and hors de combat indicators — a question that the legal review of AI-enabled operational systems is already required to ask. Not a new obligation, but the effectiveness condition of obligations that have bound States for decades.

A central premise of the analysis is temporal: as operational decision cycles contract below the threshold of meaningful human assessment, human supervision risks becoming nominal rather than substantive. Under such conditions, an embedded surrender-recognition safeguard may be the only mechanism capable of operating within the window in which lethal action is determined.

The paper does not argue that international law mandates any particular AI architecture, nor that autonomous or semi-autonomous systems are unlawful as a class. It advances a determinate proposition: the systematic exclusion of technically achievable safeguards relevant to surrender recognition and de-escalation engages existing duties of recognition, verification, and legal review — and therefore warrants documented justification rather than silence.

Revision note: This second revised edition (i) introduces a new Section III on the legal-review obligation of Article 36 AP I as the procedural anchor of the analysis; (ii) replaces paraphrases of the core provisions with verbatim treaty text, verified against official sources; (iii) consolidates the argument's qualifications into the Methodological Note; and (iv) adds a cross-reference to the parallel analysis in civil machine-safety law.


Cite As

Giovanni Nardacci, 'Lawful Operational Safeguards in AI Systems: Surrender Recognition, Compliance Architecture, and International Humanitarian Law' (2nd rev. edn, 2026), SSRN Abstract ID 6821838, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6821838.

Download

📄 Download PDF 📑 View on SSRN

Keywords

International Humanitarian Law · Autonomous Weapons · AI Governance · Lawful Operational Safeguards · Surrender Recognition · Meaningful Human Control · Temporal Compression · Defence Procurement · IHL Compliance Architecture · Article 36 AP I · Article 41 AP I · Article 57 AP I


Related Work

The temporal structure analysed in this paper is not unique to IHL. For the parallel argument in civil machine-safety law, see: Systemic Arrestability: Operator Protection in Machine Safety Law When Machine Speed Exceeds Human Reaction (2026), SSRN Abstract ID 6923461.


This paper proposes an analytical framework grounded in treaty text; it does not purport to resolve disputed questions of treaty interpretation, nor to establish binding standards of conduct beyond those the cited provisions themselves impose.

Human Flag Association — Bellinzona, Switzerland