news SIPRI 2x shared

Mapping the military AI industry | SIPRI

The SIPRI backgrounder explains that industry involvement is increasingly seen as essential to efforts to govern the responsible military use of AI, including UN-led initiatives and the REAIM summit. States want industry participation for two main reasons: companies have the technical expertise needed to understand the technology, and design choices made during AI development can affect whether military use complies with legal and ethical rules. The report argues that “industry” is difficult to define because AI is not a single technology but an ecosystem. To clarify this, it uses the idea of a military AI “stack” with three layers. The bottom layer is hardware and infrastructure, including chips, data centres, cloud services, fibre-optic cables, energy grids and networks. The middle layer is foundational software, such as foundation models, coding frameworks and data pipelines, which shape the safety and reliability of downstream systems. The top layer is applications, where AI is integrated into military systems for tasks such as targeting support, autonomous weapons, logistics, maintenance and command and control. The report also identifies four main types of actors: defence primes, which are established arms companies; neoprimes and defence start-ups, which focus on software, data and AI-enabled tools; big tech firms, which provide core digital infrastructure and sometimes military applications; and foundation model providers. The military AI industry is fast-changing, with acquisitions and partnerships between legacy defence firms and newer AI companies becoming increasingly common.

Build this article’s timeline

Extract dated events, key actors, and open questions using the same timeline analysis available in Signal.

Log in to build

Timeline generation is available to signed-in community members.

Five Ws 🤖 4 AI prefills · 1 still unanswered
Who
António Guterres, Alexander Blanchard, Vincent Boulanin, and SIPRI
🤖 AI prefill
What
The SIPRI backgrounder explains that industry involvement is increasingly seen as essential to efforts to govern the responsible military use of AI, including UN-led initiatives and the REAIM summit.
🤖 AI prefill
When When did this happen, or what is the timeframe?
Where
United States, Sweden, and France
🤖 AI prefill
Why
Centers on military AI industry, AI stack, and foundation models
🤖 AI prefill

Login to answer and earn XP for each gap you fill.

Industry Involvement industry, stakeholders, arms control, policy initiatives, responsible AI Military AI Stack stack, hardware, infrastructure, cloud, foundation models Defense Industry Actors defence primes, neoprimes, start-ups, big tech, foundation model providers Military AI Applications autonomous weapons, targeting, battle-management, logistics, maintenance
military AI industry100% AI stack98% foundation models96% application layer93% defence primes91% neoprimes89% big tech87% test, evaluation, verification and validation (TEVV)86% command and control84% autonomous weapons83%
6
Manipulation Score: Low How much this article uses emotional framing vs. straight reporting
Loaded Language 525%
Absolutist 415%
Threat/Panic 625%
Us vs Them 215%
Engagement Bait 820%
Manipulation score look right?

Factual Claims (10)

medium geopolitical
The involvement of industry is a key aspiration of arms control initiatives on responsible military AI.
medium social
There is no representative body for the AI industry, and the concept of an AI industry is not straightforward.
medium technology
The military AI stack has three layers: hardware/infrastructure, foundational software/programming, and applications.
low technology
Examples of infrastructure include data centres, fibre-optic cables, energy grids, and cloud services like Microsoft Azure or AWS.
medium technology
Foundation models can be vulnerable to prompt injections that may manipulate downstream military AI outputs.
medium technology
Palantir is considered a key player in the current military AI wave despite not focusing on building AI models.
medium military
Military-specific AI applications include targeting support, autonomous weapons, logistics, maintenance, and command and control.
medium economic
Militaries rarely build an entirely separate AI stack from scratch because the required infrastructure is prohibitively expensive.
medium technology
Claude Gov was customized for US national security customers to be used in classified environments.
low economic
Saab acquired CrowdAI in 2023 and Safran acquired Preligens in 2024.

Comments

Login to view and post comments.

🏠Portal 📰Links ❓Q&A 📅Events 💼Jobs