NATO Principles of Responsible Use of AI

What they are

NATO’s 22 October 2021 summary of its Artificial Intelligence Strategy places the NATO Principles of Responsible Use for AI in Defence at the forefront of that strategy. The principles do not affect or supersede existing obligations and commitments, both national and international. They apply across all types of AI applications and are a baseline for Allies as they use AI in the context of defence and security.

The revised AI Strategy of 10 July 2024 restates that Allies endorsed six Principles of Responsible Use (PRUs) for AI in Defence, i.e. Lawfulness, Responsibility and Accountability, Explainability and Traceability, Reliability, Governability and Bias Mitigation. NATO’s news release the same day says a priority is advancing the implementation of NATO’s Principles of Responsible Use.

The six principles (2021 wording)

Allies and NATO commit that AI applications they develop and consider for deployment will be, at the various stages of their lifecycles, in accordance with:

1. Lawfulness — developed and used in accordance with national and international law, including international humanitarian law and human rights law, as applicable.
2. Responsibility and Accountability — developed and used with appropriate levels of judgment and care; clear human responsibility shall apply in order to ensure accountability.
3. Explainability and Traceability — appropriately understandable and transparent, including through the use of review methodologies, sources, and procedures, including verification, assessment and validation mechanisms at either a NATO and/or national level.
4. Reliability — explicit, well-defined use cases, with safety, security, and robustness … subject to testing and assurance within those use cases across their entire life cycle.
5. Governability — developed and used according to intended functions, allowing appropriate human-machine interaction, the ability to detect and avoid unintended consequences, and disengagement or deactivation … when such systems demonstrate unintended behaviour.
6. Bias Mitigation — Proactive steps will be taken to minimise any unintended bias in the development and use of AI applications and in data sets.

Operationalising them

The Data and Artificial Intelligence Review Board (DARB) is the focal point for NATO’s efforts to govern responsible development and use of AI by helping operationalise PRUs. The 2022 text says the DARB’s first line of effort is to produce RAI toolkits and that a key deliverable is developing the RAI certification standard, described there as a process standard. We have not found a later official NATO page on the sources list that publishes that certification standard as a finished public specification, so we do not treat it as issued.

The 2024 revised strategy calls for key elements of an Alliance-wide AI Testing, Evaluation, Verification & Validation (TEV&V) landscape and says Allied testing facilities need to be able to determine if AI applications can be used safely and in accordance with NATO’s PRUs.

What they are not

The principles are an Allied baseline, not a national contract clause and not a product-certification method. National documents such as DoD Responsible AI and UK MOD JSP 936 remain the sources for those governments’ own requirements. NATO says some Allies already have national principles of responsible use.

Primary sources

Read the 2021 strategy summary and the 2024 revised-strategy summary on nato.int.