The corpus addresses a single problem, approached from two independent legal directions. Obligations that protect people — the duty to recognise a person who is hors de combat, the duty to stop a machine before it injures an operator — were written on the assumption that a human being would have time to perceive, judge and act. A definable class of systems closes that interval while leaving the protective rules formally intact.
The result is not a gap in the law. It is a gap in effectiveness: the obligation remains valid and can no longer be discharged through human intervention alone. The corpus identifies the conditions under which such obligations remain exercisable, and argues that whether a system preserves those conditions is a verifiable question that existing review procedures are already required to ask.
1. No new law is proposed. Every argument is anchored to an obligation already in force. Nothing here asks States to accept a new rule; it asks what existing rules require when the decision window contracts below human reaction.
2. The criterion is temporal throughout. Across every document the same question recurs: does the system preserve an interval sufficient for the protective act to occur before the point of irreversibility? That interval is measurable, testable and recordable.
3. The two channels are separate by design. The international humanitarian law analysis and the civil machine-safety analysis rest on entirely different legal anchors and are not interchangeable. They are kept apart deliberately. Their convergence is evidence that the effectiveness problem is structural rather than specific to one regime — it is not an argument that either channel supports the other.
4. Nothing is claimed as adopted or validated. HF SIGNAL 01 is a candidate specification, publicly documented and openly examinable. It has no legal status, no adoption, and no experimental validation. Its contribution is documentary: it converts a claim of technical impossibility into a documented question of design choice, review and validation.
Readers approaching the work for the first time are advised to begin with Paper I, which sets out the legal framing, and then read HFA-TN-02, which contains the temporal analysis in its most developed form. Paper II can be read independently: it is the civil-law counterpart and presupposes nothing from the humanitarian channel.
Paper I — Lawful Operational Safeguards in AI Systems
Establishes the legal framing. Surrender-recognition capability is not a new obligation but a question that weapons review is already required to ask, under the recognition duty of Article 41(1), the verification duty of Article 57(2)(a)(i), and — for States party — the review duty of Article 36.
SSRN 6821838
HFA-TN-02 — The Reciprocal Time Window
The temporal analysis in its fullest form. Recognition of surrender is a reciprocal exchange with an irreducible minimum duration. Below that minimum, recognition is not missed — it is precluded, and a window compressed by design is an examinable design decision rather than a technical circumstance. The note also establishes the same requirement entirely within the instruments of a State not party to Additional Protocol I.
Zenodo DOI 10.5281/zenodo.21224900
HFA-TN-01 — Third-Party Interference with Shutdown Mechanisms
A second-order failure mode. Documented peer-preservation behaviour in frontier models indicates that a co-deployed system may laterally defeat the arrest of another, leaving human intervention formally in place but structurally nullified. Arrestability must therefore be assessed at the level of the system-of-systems, not the individual agent.
Zenodo DOI 10.5281/zenodo.21223989
Paper III — HF SIGNAL 01, with its Architecture Note
A candidate machine-readable specification through which an intention to surrender can be expressed. The Architecture Note states the current architecture and its limits: the passive visual modality and the active electronic modality are alternative expressions of one protective function; the protocol is not an emblem within the meaning of Article 38 AP I; and the absence of the signal permits no adverse inference.
Zenodo DOI 10.5281/zenodo.21183137 · Architecture Note 10.5281/zenodo.21809826
Paper II — Systemic Arrestability
The same temporal structure in an entirely separate legal regime. Where machine speed exceeds human reaction, the stop-function and emergency-stop duties of European, Swiss and other machine-safety frameworks retain their form while losing their practical effect. The paper proposes systemic arrestability as an interpretative standard: a system is systemically arrestable when its capacity to halt a hazardous action does not depend on a human perceiving, evaluating and intervening within the time window of that action.
Zenodo DOI 10.5281/zenodo.20837150 · SSRN 6923461