Governments are increasingly discussing how to build legal and policy guardrails for military artificial intelligence, but the more immediate question is how existing tools can already shape its development. One of the most powerful, yet often overlooked, is defence procurement.
The argument advanced by Dr Anke Allenhöfer in Articles of War is that procurement should not be treated as a narrow purchasing exercise. It is also a governance mechanism, capable of steering how mili...
Continue Reading This Article
Enjoy this article as well as all of our content, including reports, news, tips and more.
By registering or signing into your SRM Today account, you agree to SRM Today's Terms of Use and consent to the processing of your personal information as described in our Privacy Policy.
tary AI is designed long before it is fielded. By setting operational requirements, technical specifications, evaluation criteria and contractual obligations, governments can influence the choices made by defence companies at the stage where software architecture, training data, testing methods and human-oversight features are still being determined.
That matters because many of the most important compliance questions arise during design rather than deployment. International humanitarian law sets the rules governing distinction, proportionality and precautions in attack, but it does not prescribe the engineering solutions needed to meet them. Article 36 of Additional Protocol I requires States to review new weapons, means and methods of warfare before study, development, acquisition or adoption, yet the practical content of those reviews has become more difficult as AI systems grow more complex. Related academic work has argued that AI decision-support systems may also need to be brought within the scope of review, and that a lifecycle-based approach is increasingly necessary.
Allenhöfer’s point is that procurement can translate those legal obligations into concrete requirements. Mandatory eligibility criteria can be used to exclude systems that do not satisfy verifiable IHL standards, while weighted criteria can reward qualities such as explainability, traceability, robustness and auditability. In effect, governments can use tendering to signal which design choices will be commercially valued, nudging industry towards safer and more defensible systems.
The article also argues that procurement should not stop at the award stage. AI-enabled military systems evolve over time through updates, retraining and adaptation to new environments, which means legal review cannot sensibly be a one-off exercise. Contracts can require repeated review, testing and modification when a system changes materially, turning compliance into an ongoing obligation rather than a single gateway check.
That lifecycle approach reflects wider concerns in the literature. A recent systematic review of ethical governance in defence AI identified persistent barriers to implementation, including rapid technological change, a lack of standardisation and weak interdisciplinary collaboration. Those findings help explain why governance frameworks often look stronger on paper than in practice. If legal, ethical and technical teams remain isolated from one another, the result is often retrospective scrutiny rather than design-stage control.
Allenhöfer therefore proposes a more integrated model inside defence companies as well as outside them. Her concept of Integrated Legal-Ethical Engineering Governance brings together legal advisers, compliance specialists, engineers and ethics teams at fixed checkpoints during development, so that concerns can be identified before design decisions harden into expensive commitments. The broader claim is that responsible military AI depends not only on external rules, but on internal organisational structures capable of translating those rules into engineering practice.
The article stops short of suggesting that procurement can replace formal regulation. International and national standards remain essential if States are to create durable and uniform rules. But procurement offers something different: immediate leverage. Before new treaties are agreed, and alongside them afterwards, governments already control purchasing power large enough to shape what the market builds.
In that sense, the article’s central message is pragmatic rather than aspirational. States need not wait for a new global framework before influencing responsible military AI. They already possess a tool that can do so, if they choose to use it as more than an administrative formality.
Source: Noah Wire Services
