NATO’s Data and Artificial Intelligence Review Board

Introduction

1. In the NATO Artificial Intelligence (AI) Strategy, Allies and NATO committed to ensuring that the AI applications they develop and consider for deployment will be in accordance with six Principles of Responsible Use (PRUs): lawfulness; responsibility and accountability; explainability and traceability; reliability; governability; and bias mitigation.
    
2. The Data and Artificial Intelligence Review Board (DARB) serves as a forum for Allies and the focal point for NATO's efforts to govern responsible development and use of AI by helping operationalise PRUs.
    
3. The DARB is being established with the following aims:
   • Building Trust – with publics, innovators and operational end users – as well as within the international community to steer responsible defence innovation efforts in accordance with our values, norms, and international law;
   • Guiding Responsible AI (RAI) Adoption – by translating PRUs into user-friendly RAI standards and best practices that offer the NATO Enterprise and Allies a common baseline to help create quality controls, mitigate risks and adopt trustworthy and interoperable AI systems; and
   • Acting as a Forum – for Allies and the NATO Enterprise to share best practices, and exchange views.
4. The DARB must ensure that its outputs do not raise unnecessary barriers to adopting AI at the speed of relevance. Through the DARB, Allies and NATO will hone RAI practices that deliver more reliable, interoperable and safer systems, helping deliver a qualitative advantage relative to strategic competitors and potential adversaries.

Functions

5. The DARB will guide RAI implementation in an agile way that both anticipates and adapts to changing circumstances and technological advancements. The DARB will aim to achieve this by creating practical RAI toolkits, guiding RAI implementation in NATO, and supporting Allies in their RAI efforts.

Creating Practical RAI Toolkits

6. The DARB's first line of effort will be to produce RAI toolkits for use by the NATO Enterprise, which will also be available to those Allies who wish to use them. One of the DARB's key deliverables in this line of work is developing the RAI certification standard. The RAI certification standard will be a process standard, building on experience from use cases, inputs from NATO Enterprise stakeholders, and best practices outside of the NATO context, including from public, private and academic sectors, as well as civil society and other international actors.
    
7. In the production of these toolkits, the DARB and its subgroups will sustain engagement with a broad range of stakeholders from within defence and security sectors, as well as civilian-oriented RAI initiatives.

Guiding RAI Implementation in NATO

8. The DARB will guide RAI implementation across the NATO Enterprise. AI and data exploitation use case leads and project owners will periodically complete self-assessments on adherence to the PRUs. NATO Enterprise stakeholders will ensure compliance with the RAI certification standard, once issued.

Supporting Allies in their RAI Efforts

9. At the national level, Allies may choose to use RAI toolkits and implement responsible-by-design practices recommended by the DARB. To facilitate this function, the DARB will collect and share information stemming from the aforementioned functions with Allies, for their consideration to take up on a voluntary basis in national efforts.

Composition

10. Each NATO nation will nominate one national nominee to serve on the DARB. National nominees may be drawn from government, academia, the private sector and civil society, with AI expertise in various relevant fields – such as computer science, data analytics, engineering, humanities, law, philosophy and social sciences.