The following projects constitute a sovereign AI stack designed to resist centralized regulation and legal suppression by distributing data, compute, coordination, and finance across decentralized networks.
Companies
Mistral AI
A French firm specializing in high-efficiency models like Mistral Large 3. They serve as a European vanguard for the AI freedom movement, resisting attempts to impose strict liability on open-source developers [1].DeepSeek AI
This Hangzhou-based firm disrupted the industry in 2025 with the release of DeepSeek-V3 and DeepSeek-R1. R1 matches the reasoning capabilities of leading proprietary models while being released under a permissive MIT license, effectively commoditizing high-tier reasoning for the global developer community [2].Alibaba Cloud (Qwen)
The Qwen team released the Qwen 3 series in April 2025, featuring a 235B Mixture-of-Experts (MoE) flagship. With native support for 119 languages and a 1-million-token context window, Qwen 3 provides a robust, non-Western alternative for global AI infrastructure [3].Meta (Llama)
The Llama 4 family, including the Maverick (400B) and Scout models, remains the industry standard for open-weight intelligence. By providing weight downloads for models reaching up to 2 trillion parameters (Behemoth), Meta enables local deployment at a scale previously reserved for the largest cloud providers [4].Moonshot AI
Released Kimi K2 in July 2025, a massive open-source model with approximately 1 trillion parameters. This demonstrates that open-weights systems can match the scale of the largest proprietary architectures [5].Hugging Face
A central hub hosting hundreds of thousands of models. Its leadership advocates for open science as a fundamental right, opposing hardware-level monitoring and restrictive licensing [6].Anyscale
Creators of the Ray framework, facilitating distributed computing that allows researchers to train models across fragmented hardware, bypassing centralized data centers.Baseten
Provides high-performance inference for custom models, enabling enterprises to deploy AI independently of proprietary APIs.
Open Source Software & Protocols
Monero (XMR)
A cryptocurrency built for mandatory privacy using ring signatures and stealth addresses. In the AI stack, Monero serves as the primary tool for censorship-resistant funding, allowing researchers to receive payment and purchase compute power without being subject to bank-level “de-banking” [7].Zano
A privacy-first blockchain that combines anonymity with smart contract capabilities. It supports Confidential Assets, allowing the creation of private stablecoins and decentralized marketplaces where AI services, data, and compute can be traded in an encrypted environment [8].Zcash (ZEC)
The first major blockchain to implement zk-SNARKs (Halo 2), providing privacy for shielded transactions. Its high-performance cryptography allows for verifiable but private financial coordination within the sovereign AI ecosystem [9].Arweave
A decentralized “permaweb” providing immutable storage. Once training weights are uploaded, they cannot be deleted, ensuring model “immortality” regardless of regulatory changes.Akash Network
A supercloud marketplace for GPU power. Its permissionless environment allows developers to migrate workloads across jurisdictions to bypass localized killswitches [7].Morpheus
A protocol connecting inference providers with users via smart contracts, making compulsory licensing schemes technically impractical to enforce.Common Crawl
An open-source web archive serving as primary training data. Widespread mirroring creates “shadow data” that remains accessible even under new scraping restrictions.Darkfi
A Layer 1 blockchain using zero-knowledge cryptography to allow developers to collaborate and receive funding without revealing legal identities, protecting them from personal legal targeting.Bittensor
A marketplace protocol that distributes model training across a global network of subnets. Known as the Bitcoin of AI, it shifts compute to permissive jurisdictions if one region imposes restrictions [9].Nillion
A network utilizing Nil Message Compute to process data without nodes ever seeing it, preventing hardware-level monitoring of AI training.Gensyn
Connects consumer hardware into a global supercomputer using probabilistic proofs to verify training, removing the need for central auditors.Artificial Superintelligence Alliance
A framework for autonomous economic agents that can independently transact for compute and data, existing outside traditional legal structures [7].Internet Computer (ICP) (On-Chain AI)
Runs AI models entirely on-chain as tamper-proof smart contracts, eliminating platform risk by removing reliance on centralized cloud providers.
IntuitionLabs, Mistral Large 3: An Open-Source MoE LLM Explained, IntuitionLabs, 2025. This article explains the technical advancements and the Apache 2.0 licensing of Mistral’s flagship models released to ensure developer autonomy. https://intuitionlabs.ai/articles/mistral-large-3-moe-llm-explained
BentoML, The Complete Guide to DeepSeek Models: V3, R1, V3.1, V3.2 and Beyond, BentoML, 2025. This comprehensive guide details the performance and open-source licensing of the DeepSeek-R1 series, which brought frontier-level reasoning to the public domain. https://www.bentoml.com/blog/the-complete-guide-to-deepseek-models-from-v3-to-r1-and-beyond
Apidog, Best Qwen Models in 2025, Apidog, 2025. This review analyzes the Qwen 3 release, highlighting its massive parameter scale and its role as a leading open-weights alternative to US-based frontier models. https://apidog.com/blog/best-qwen-models/
Skywork AI, Llama 4, 2 Trillion Parameters & Open-Weight AI: The 2025 Builder’s Shift, Skywork AI, 2025. The report discusses Meta’s 2025 release of Llama 4, focusing on the 2T parameter Behemoth model and its impact on large-scale self-hosted AI. https://skywork.ai/blog/llama-4-open-weight-2025/
HPCwire, China’s Moonshot AI Releases Trillion Parameter Model Kimi K2, HPCwire, 2025. This report details the release of the Kimi K2 model, highlighting its scale and strategic use of open-weights to match proprietary performance. https://www.hpcwire.com/2025/07/16/chinas-moonshot-ai-releases-trillion-parameter-model-kimi-k2/
The Key Executives, Building AI for Everyone: Clément Delangue’s Open-Source Revolution at Hugging Face, The Key Executives, 2025. This profile examines the advocacy for transparent and community-driven AI development as a safeguard against centralized control. https://www.thekeyexecutives.com/2025/03/12/building-ai-for-everyone-clement-delangues-open-source-revolution-at-hugging-face/
CCN, Artificial Superintelligence Alliance Price Prediction 2025, CCN, 2025. This analysis discusses the market positioning of the ASI merger and the role of privacy coins and decentralized compute in the broader sovereignty thesis. https://www.ccn.com/analysis/crypto/asi-price-prediction-2024-fet-price-analysis/
CoinMarketCap, What Is Zano (ZANO) And How Does It Work?, CoinMarketCap, 2026. A technical deep dive into Zano’s hybrid consensus and confidential assets which enable private decentralized applications for AI resource trading. https://coinmarketcap.com/cmc-ai/zano/what-is/
Galaxy, Why Has Zcash Suddenly Soared?, Galaxy Research, 2025. This research piece outlines Zcash’s 2025 resurgence and how its zk-SNARK infrastructure provides a mathematically sound alternative to monitored financial systems. https://www.galaxy.com/insights/research/zcash-price-zec-near-intents-zashi-wallet-privacy-zero-knowledge-proofs
Want to stay in touch?
Join my Signal announce-only group to be notified when I have a new essay up and other important announcements.
For discusion, join the libertygardeners Signal group.
Subscribe to my mailing list.
Email: archerships@protonmail.ch
Signal: archerships.43
Other social media: