Building Decentralised AI: Democratic Alternatives¶

Why this page exists¶

Even technically safe AGI creates governance challenges: Who controls transformative technology, how are benefits distributed and what alternatives exist when a provider fails? This page explores decentralisation and democratic control as one approach that complements frontier safety, international coordination, national policy and the C·A·G·R framework.

The problem: Power concentration¶

Australian government and critical infrastructure could depend on a handful of foreign AI providers, giving those providers leverage over data, service continuity and sovereign decision-making. The issue is not only whether systems work, but who controls them, who benefits and whose values they serve.

→ See how this unfolds: Power Concentration & Governance Failure

The d/acc approach: Differential acceleration toward defence and democracy¶

This page draws on Vitalik Buterin's d/acc (defensive/decentralisation/democratic accelerationism): a pro-building approach that prioritises defensive capability, transparency and democratic oversight over capability scaling for its own sake.

D/acc emphasises four interconnected principles:

1. Differential acceleration¶

Speed up community infrastructure, open alternatives and governance tools that make society resilient—not only systems that increase capability.

2. Defensive advantage¶

Make defence easier than attack through transparency, local deployment, exit rights and redundancy.

3. Decentralised systems¶

Build community-owned compute, open models and cooperative governance without a single controlling chokepoint.

4. Democratic participation¶

Give members and affected communities meaningful ownership and decision rights, not consultation alone.

D/acc aims to increase coordination while limiting domination. It is one proposed approach to containing risks early while distributing access and decision-making.

How d/acc supports C·A·G·R¶

D/acc isn't a replacement for the C·A·G·R framework—it's a set of principles for implementing parts of it:

• Containment: Defensive measures can make systems harder to compromise; distributed evaluation may broaden safety oversight
• Alignment: Democratic participation can bring community values into decisions alongside technical and organisational objectives
• Governance: Decentralisation and democratic participation can distribute authority rather than relying only on regulation of concentrated power
• Resilience: D/acc principles can support resilience through redundancy, transparency, exit rights and distributed capability

What decentralised AGI infrastructure looks like¶

Decentralisation isn't just theory. Relevant infrastructure and governance models exist across four layers:

Technical layer: Open models and local compute¶

Open-weight models such as Llama, Mistral, Qwen and DeepSeek allow organisations with sufficient hardware and expertise to run capable AI locally. Important projects include:

• Hugging Face — A major open model, dataset and deployment hub
• Ollama — Run models locally with a single command; makes local AI accessible to non-experts
• vLLM and llama.cpp — Efficient inference engines that reduce hardware requirements
• Together AI and Fireworks — Independent inference providers offering alternatives to frontier-lab APIs
• Decentralised training — Peer-to-peer GPU networks enable training runs distributed across globally sourced hardware, without passing through centralised cloud providers (Kryś, Sharma & Egan 2025). This has governance implications: decentralised training could eventually undermine compute-based governance by routing around regulated infrastructure.

Why this matters: Local deployment can give you more control over data handling, service continuity and costs. It can also reduce exposure to API withdrawal, changing provider terms and foreign-jurisdiction access, although local operation introduces its own security, maintenance and governance responsibilities.

Coordination layer: Crypto and verification infrastructure¶

Blockchain and cryptographic tools can support coordination without relying on a single central authority:

• World — A proof-of-personhood approach intended to distinguish humans from AI agents at scale. Biometric identity systems also create substantial privacy, consent and regulatory risks; Spain's data protection authority, for example, ordered a precautionary halt to Worldcoin data processing in 2024.
• Prediction markets (Polymarket, Metaculus) — Aggregate distributed knowledge about AI timelines and risks
• Gitcoin — Quadratic funding for public goods, including AI safety research
• DAOs (decentralised autonomous organisations) — On-chain mechanisms for collective decision-making, with governance and legal limitations that vary by design and jurisdiction

Verification and transparency:

• zkML — Zero-knowledge proofs that verify AI model outputs without revealing inputs or weights
• C2PA — Content authenticity standards to verify AI-generated vs human content
• Cryptographic commitments — Prove what model produced an output, enabling accountability

Why this matters: As AI agents become more common, organisations may need better ways to verify identity, coordinate resources and record collective decisions. Cryptographic mechanisms offer some options, but their suitability depends on governance, security, privacy, accessibility and legal context.

Ownership layer: Cooperatives and community infrastructure¶

Platform cooperatives apply democratic ownership to AI infrastructure:

• Stocksy — Artist-owned stock photography cooperative (model for AI training data ownership)
• Resonate — Music streaming cooperative (model for AI service delivery)
• Driver's Seat Cooperative — Historical gig-worker data cooperative (a model for collective bargaining with AI platforms)

Compute cooperatives pool resources for shared AI infrastructure:

• Community GPU clusters funded by member contributions
• Shared fine-tuning and hosting costs across organisations
• Collective bargaining power with cloud providers

Australian precedents: Credit unions (member-owned banking), energy cooperatives (community-owned solar/wind), agricultural cooperatives (collective marketing). These models can extend to AI compute, data and services.

Governance layer: Democratic decision-making tools¶

Participatory governance where users genuinely control AI systems:

• Decidim — Open-source platform for participatory democracy, used by Barcelona, Helsinki and others
• Pol.is — AI-assisted consensus-finding used in Taiwan's vTaiwan process
• Loomio — Collaborative decision-making for organisations
• Snapshot — Off-chain governance voting for DAOs and communities

Real examples of democratic tech governance:

• Taiwan's g0v movement — Civic technologists building open-source alternatives to government services; vTaiwan used Pol.is to achieve consensus on Uber regulation in 2015-2016
• Barcelona's digital sovereignty — City runs on open-source, uses Decidim for participatory budgeting, prioritises technological sovereignty over Big Tech lock-in
• Mondragon Corporation (Spain) — 70,000+ workers in a federation of cooperatives, demonstrating democratic governance at industrial scale

Why this matters: Technical decentralisation does not automatically distribute authority. Democratic governance tools can give communities—not only developers or early investors—a meaningful role in decisions about how AI systems are owned and used.

Can decentralised approaches actually compete?¶

Different tools suit different purposes. Centralised AI currently excels at frontier capability and massive compute; decentralised approaches can offer privacy, local customisation, democratic governance, resilience and exit rights. The useful question is not whether every local system can match the frontier, but where alternatives provide distinct value.

Unique advantages of decentralised AI:

• Operate during disruption — local systems work when cloud services fail
• Reflect local values — communities govern to match their values, not corporate objectives
• Maintain genuine privacy — data stays within your infrastructure
• Enable exit rights — switch providers without losing everything
• Experiment with governance — different communities try different models

Centralised and decentralised approaches can coexist. The strategic risk is monoculture: without viable alternatives, users have little leverage when providers fail or change terms.

Why Australia is well-positioned¶

Existing foundations: Cooperative organisations in banking, energy and agriculture; established democratic institutions; renewable-energy resources; and universities with relevant technical, legal and policy expertise.

Strategic need: Geographic isolation and a relatively small domestic market can increase dependence on foreign supply chains, platforms and AI providers. Decentralised or cooperative options may reduce some concentration risks when they provide credible alternatives, portability and local control.

Theory of change: How do we get there?¶

The aim is not to replace frontier providers immediately, but to create negotiating power through exit rights and demonstrate viable alternatives.

The phases below are an illustrative sequence, not a forecast or government programme.

Phase 1: Test viability (near term) — Pilot local AI, shared compute, community procurement and cooperative governance; build practical deployment and legal capability.

Phase 2: Build scale and legitimacy — Expand approaches that perform well, strengthen interoperability and give organisations credible exit options.

Phase 3: Sustain a mixed ecosystem — Support multiple viable models: centralised coordination, decentralised local control and cooperative ownership of shared resources.

The urgency: Why act now¶

Once provider lock-in, network effects and incumbent standards deepen, alternatives become harder to establish. Current conditions still leave room to shape the landscape:

• Some open-weight AI models are suitable for practical local uses
• Community infrastructure and cooperative-governance experiments provide models to test
• Organisations can still negotiate portability and exit rights before dependencies deepen
• Policy and standards are still developing

These are reasons to test alternatives now—not evidence that decentralisation will inevitably succeed or that every system should be decentralised.

Where to next¶

• C·A·G·R Framework — comprehensive approach to AGI preparation
• AGI Scenarios — how power concentration plays out in practice
• Power Concentration Scenario — detailed exploration of governance failure
• Key Concepts & Glossary — definitions of terms used on this page
• External Resources — organisations and tools for building alternatives

Common questions¶

"Won't market competition solve this? Why does government need to intervene?"

Frontier training requires enormous capital, while network effects and switching costs favour incumbents. Competition helps only when users have viable alternatives and can move between them. Strategic sectors may therefore need interoperability, procurement standards or public investment to preserve choice.

"Won't this slow down AI development?"

Some centralised development may be faster, but monoculture is fragile. Diverse approaches can improve robustness and experimentation. The strategic question remains: What speed, for whom, toward what ends?

"Can cooperatives really compete with Big Tech?"

They can compete where privacy, local customisation, trust or democratic ownership matter, even if they do not lead at the frontier. The objective is genuine choice and provider accountability, not replacing every centralised service.

"Is Australia too small for this?"

Achieving scale alone is difficult. Regional cooperation with Pacific and Southeast Asian partners could pool capability while retaining local governance. Australia's cooperative tradition and research base provide useful foundations.