# Meta Defense Meta's defense and national security portfolio — spanning open-source AI for military operations, XR combat systems, surveillance infrastructure, and intelligence platform access. Unlike [[Google Defense]] or [[Palantir]], Meta didn't enter defense through direct government contracts. It arrived through three vectors: open-sourcing powerful AI models, building the world's largest personal data platform, and partnering with existing defense primes. As of 2025, Meta is officially a defense contractor. ## Current Defense Portfolio ### 1. Llama Military AI Platform The centerpiece. See: [[Llama Military Adoption]] for full details. - **November 2024**: Meta lifted its acceptable use policy ban on military applications of Llama, explicitly granting access to U.S. government agencies and defense contractors. - **Partners**: [[Lockheed Martin]], Booz Allen Hamilton, [[Anduril]], [[Palantir]], [[Scale AI]], Accenture, Deloitte, IBM, Leidos, Microsoft, Oracle, Amazon Web Services, [[Databricks]], Snowflake. - **Deployments**: Dozens of military and intelligence organizations using Llama by September 2025. AWS and Snowflake expanding access to classified networks. Five Eyes nations → NATO allies. - **Use cases**: Intelligence analysis, code generation, data processing, operational planning, logistics optimization. - **Key distinction**: Unlike Google (Gemini) or [[OpenAI]] (ChatGPT), Meta's models are **open-source and open-weight** — meaning military users can fine-tune, modify, and deploy them on air-gapped classified networks without Meta's involvement. This is both the strategic advantage and the proliferation risk. ### 2. Anduril-Meta XR Combat Systems Meta's highest-profile defense partnership, announced May 2025: - **Eagle Eye**: Military XR helmet system built on Meta Reality Labs hardware, integrating Llama AI and [[Anduril Lattice]] command-and-control platform. Designed as a complete helmet replacement with night vision, thermal sensing, and AR battlefield intelligence overlay. - **SBMC Next (formerly IVAS Next)**: The Army stripped the $21.9B IVAS program from Microsoft in February 2025 and re-competed it. Anduril-Meta team selected alongside newcomer Rivet (September 2025). The system allows soldiers to control autonomous platforms and access real-time combat intelligence through XR. - **Technology flow**: Meta provides the XR hardware and AI models. Anduril provides defense integration, Lattice C2, and military-grade ruggedization. [[Palmer Luckey]]'s journey — from Oculus founder → fired by Zuckerberg → Anduril founder → back to building military hardware with Meta technology — is the narrative arc of Big Tech's defense pivot. ### 3. Surveillance Infrastructure (Legacy) Meta didn't intend to build military surveillance tools. It built an advertising platform that happens to be the most comprehensive human behavioral database ever created: **Data Scale**: - 3 billion+ daily active users across Facebook, [[Instagram]], [[WhatsApp]] - Meta's AI processes 3 billion+ pieces of content daily - Location data, social graphs, messaging metadata, behavioral patterns, facial images **Known Law Enforcement/Intelligence Use**: - **Geofeedia (2016)**: ~500 police departments used Facebook/Instagram data via Geofeedia to monitor [[Black Lives Matter]] protests. Geofeedia labeled activists and unions as "overt threats." Meta cut access after ACLU exposure. - **[[Cambridge Analytica]]/SCL Group (2014–2018)**: 87 million Facebook profiles harvested. Parent company SCL Group was a UK defense/intelligence contractor with List X top-secret clearance. Ran psyops for British Army (15 Psyops Group), NATO, [[Saudi Arabia]], Norwegian defense, and U.S. agencies. Facebook data → military psychological operations. Key SCL advisors included [[Michael Flynn]] and [[John Bolton]]. - **[[Clearview AI]]**: Scraped 20B+ images from Meta platforms for facial recognition database used by 2,000+ law enforcement agencies, FBI, and ICE. - **Government data requests**: Meta complies with hundreds of thousands of government data requests annually (exact numbers in Meta Transparency Reports). WhatsApp shares metadata with DOJ, UK, Brazil, India despite end-to-end encryption. - **Traffic analysis vulnerability**: WhatsApp security team warned internally (2024) that governments were using traffic analysis to bypass encryption and determine who communicates with whom, group memberships, and possibly locations. Meta called it "theoretical." ### 4. Ray-Ban Meta Smart Glasses Consumer wearable with significant surveillance implications: - **Scale**: 7M+ units sold by 2025. - **Law enforcement adoption**: ICE agents reportedly wearing them in the field. CBP agents documented using them during immigration enforcement operations. - **Facial recognition risk**: No built-in facial recognition, but Harvard researchers demonstrated combining the glasses with off-the-shelf tools for near-real-time identification of strangers. - **Privacy failures**: Recording LED can be covertly disabled. Contractors in Kenya reviewing user footage — "We see everything, from living rooms to naked bodies." Users cannot opt out of the data pipeline. - **Legal exposure**: Class action lawsuit filed March 2026 alleging Meta violated federal and state privacy laws by transmitting footage to servers and subcontractors without adequate disclosure. - **EFF warning (March 2026)**: Electronic Frontier Foundation issued explicit advisory against purchasing the glasses. ### 5. Content Moderation AI → Dual-Use Capability Meta's content moderation systems — detecting terrorism content, violence, misinformation, and coordinated inauthentic behavior across billions of posts — represent one of the world's most advanced threat detection AI capabilities: - Transformer models for text analysis - Convolutional neural networks for image analysis - Graph neural networks for social network/behavioral analysis - Recurrent neural networks for temporal pattern analysis - 95% hate speech detection rate, 99% spam detection before user report These capabilities directly parallel military requirements for threat detection, intelligence analysis, information operations, and targeting. The content moderation → military AI pipeline isn't speculative — it's the same fundamental technology applied to different domains. ## What Meta Doesn't Have (Yet) Unlike [[Google Defense]] (JWCC cloud, $200M Gemini contract) or [[Palantir]] ($10B Army contract), Meta has **not been awarded large direct Pentagon contracts**. The $200M+ AI contracts went to Google, OpenAI, [[Anthropic]], and xAI — not Meta. Meta's defense play is: 1. **Indirect** — through Llama's open-source distribution to defense contractors 2. **Partnership-based** — through the Anduril XR collaboration 3. **Infrastructure-adjacent** — through the surveillance data platform 4. **Emerging** — still building toward direct government revenue This may change. The Anduril-Meta SBMC Next bid, if successful, could be worth billions. ## Lobbying & Political Spending - **2024 lobbying**: $24.4M (record high, 27% increase over 2023) - **2025 lobbying**: $26.3M (new record, 43% increase in Q1 alone) - **Q1 2024**: $7.6M as Congress advanced TikTok ban - **2024 cycle political contributions**: $5.5M - **Trump inaugural donation**: $1M - **Trump lawsuit settlement**: $25M - **Lobbying focus**: Online safety, content moderation policy, AI regulation, antitrust — not explicitly "defense" per OpenSecrets categorization, but the TikTok/national security frame overlaps significantly. ## Competitive Positioning Meta's defense strategy is fundamentally different from other Big Tech companies in this KB: - **vs. [[Google Defense]]**: Google sells cloud + proprietary AI (Gemini) directly to the Pentagon. Meta gives away AI (Llama) and partners with defense primes. Google has direct revenue; Meta has ecosystem influence. - **vs. [[Palantir]]**: Palantir sells data integration platforms to government. Meta *is* the data platform governments want access to. Palantir is a Llama deployment partner. - **vs. AWS ([[AWS Intelligence Cloud]])**: AWS sells classified cloud infrastructure. Meta's models run *on* AWS for military users. Complementary, not competitive. - **vs. [[Anduril]]**: Anduril builds hardware and autonomous systems. Meta provides the AI and XR technology inside them. Direct partnership rather than competition. ## The Strategic Logic Meta's defense portfolio makes sense through one lens: **Meta wants to be essential infrastructure for the U.S. national security state** — not by winning contracts directly, but by making Llama the default AI model for allied militaries, Meta's XR the default soldier interface, and Meta's data platforms an unavoidable component of modern intelligence. The open-source strategy means Meta doesn't need government procurement processes — it needs government *adoption*. ## See Also - [[Mark Zuckerberg]] — The man driving the pivot - [[Llama Military Adoption]] — The open-source AI strategy - [[Palmer Luckey]] — Fired by Meta, now Meta's defense partner - [[Anduril]] — Primary defense partnership - [[Peter Thiel]] — Early investor, defense ecosystem architect - [[Google Defense]] — Comparison: Big Tech direct contracting model - [[Dual-Use AI]] - [[Military-Industrial Complex 2.0]] - [[Defense Tech Conflicts of Interest]]