FLock.io is a decentralized artificial intelligence (AI) platform designed for privacy-preserving, collaborative AI model training. It utilizes federated learning and blockchain technology to enable individuals and organizations to co-create AI models without sharing their underlying private data, aiming to democratize access to AI development. ^{[1] [2]}

Overview

FLock.io was developed to address the increasing centralization of AI development within a few large technology corporations. The project's mission is to dismantle these "walled gardens" by creating an open and collaborative ecosystem where control over AI is distributed among its participants. The platform's core philosophy is built on the principles of data sovereignty and user ownership, encapsulated in its slogan, "Not your models, not your AI." By combining machine learning techniques with decentralized infrastructure, FLock seeks to solve key problems in the AI industry, including data monopolies, model biases, a lack of transparency, and the risks associated with single points of failure. ^{[2] [3]}

The platform's architecture is founded on two primary technologies. The first is Federated Learning (FL), a machine learning approach that allows AI models to be trained on decentralized data sources without the data ever leaving its local environment. This method ensures that sensitive or proprietary information remains private while still contributing to the improvement of a global model. The second is a blockchain layer, which provides the trustless foundation for coordinating tasks, managing economic incentives, and facilitating decentralized governance. This layer, built within the Base ecosystem, uses smart contracts to automate reward distribution and enforce network rules, ensuring transparency and security for all participants. The project is based in London, UK, and was founded by a team that includes computer science researchers from the University of Oxford. ^{[8] [4]}

History

The project's development roadmap outlines a phased approach to building its platform. In the second quarter of 2023, the team initiated the testnet build and launched a feature for Large Language Model (LLM) fine-tuning. Throughout the remainder of 2023, development focused on optimizing this feature, incubating use cases, and completing research into Zero-Knowledge Proofs (ZKPs) for enhanced security. ^{[3] [4]}

In 2024, the project shifted its focus to platform completion and community engagement. The AI Arena, a core component of the ecosystem, launched its closed and open beta versions in the second and third quarters, respectively. The FL Alliance, the federated learning environment, also entered its beta phase during this period. In September 2024, FLock was awarded a research grant from the Ethereum Foundation as part of its Academic Grants Round 2024 to fund research into incentive mechanisms for blockchain-based machine learning. By the end of 2024, the platform was considered feature-complete, with a focus on scaling participation and increasing the number of model calls. A key milestone in 2025 was the announcement on April 28 of "gmFLOCK," a major ecosystem upgrade designed to enhance the utility of the native FLOCK token. This mechanism aims to transform the FLOCK token from a simple tradable asset into "productive capital" by requiring users to stake FLOCK to generate gmFLOCK, which is necessary for participating in key network roles. This model is intended to create a sustainable economic flywheel for the decentralized AI ecosystem by encouraging long-term engagement and reducing speculative volatility. The project's native token, FLOCK, experienced significant price movements in 2025, recording an all-time low of $0.03517 on April 7 and reaching an all-time high of $0.6674 on September 9. The token was subsequently listed on the Upbit cryptocurrency exchange around September 12, 2025. ^{[8] [3] [5] [6]}

Technology and Architecture

FLock.io's platform is built on a two-layer architecture that integrates artificial intelligence processes with a secure blockchain foundation. This structure is designed to support a multi-stage workflow for the creation, refinement, and deployment of AI models. ^[4]

Core Components

The platform's workflow is divided into three distinct stages, each serving a specific function in the AI model lifecycle:

1. AI Arena: This is a decentralized platform for the initial training and validation of machine learning models. When a new task is created, participants known as Training Nodes compete to train models using public datasets or their own local data. Another set of participants, called Validators, then evaluate the submitted models to reach a consensus on their performance, which determines rankings and reward distribution. ^[2]
2. FL Alliance: This component provides a federated learning environment for the privacy-preserving fine-tuning of models that have been trained in the AI Arena. Participants, or FL Clients, use their private, local data to collaboratively refine a global model without the raw data ever leaving their devices. The system uses an on-chain verifiable random function (VRF) to assign roles of "proposer" (who trains the model) and "voter" (who validates updates) to prevent collusion. ^[4]
3. AI Marketplace (or AI Moonbase): This is the final deployment stage where finished models are hosted and made accessible to developers and applications. Model Hosts deploy the winning models from the previous stages, and end-users can integrate them via APIs or SDKs. Access is managed through a payment and staking system using the native FLOCK token. ^[1]

Key Technologies

The platform's functionality is enabled by a combination of specialized technologies:

• Federated Learning (FL): This is the core technology for privacy preservation. It enables collaborative model training across numerous decentralized devices without exchanging the underlying data. Only model updates, such as gradients or weights, are shared and aggregated, ensuring data sovereignty. ^[2]
• Blockchain Integration: The platform operates on the Base blockchain, which provides the infrastructure for secure and transparent coordination. It utilizes a Proof-of-Stake (PoS) consensus mechanism to manage on-chain incentives, including staking and slashing penalties, and to facilitate decentralized governance through a DAO. ^[4]
• Low-Rank Adaptation (LoRA): FLock employs LoRA, a technique that allows for the fast and compute-efficient fine-tuning of large models. This makes participation more accessible by reducing the computational resources required to contribute to model improvement. ^[1]

Security

The system is designed to mitigate a range of potential attacks common in decentralized networks. A staking requirement, where participants must lock up FLOCK tokens, makes Sybil attacks (creating multiple fake identities) prohibitively expensive. The reward system is performance-based, which discourages free-riding, as participants who do not contribute meaningfully are not compensated. To prevent model poisoning in the FL Alliance, a majority voting system for model updates is combined with a reward-and-slash mechanism to penalize and remove malicious clients. The platform's smart contracts have been audited by the security firm Slow Mist. ^{[2] [1]}

FLock API Platform

The FLock API Platform is a tool that enables developers to access and integrate AI models from the FLock marketplace into their applications. It provides serverless API endpoints, allowing for easy integration with minimal setup. The platform is designed to be compatible with the OpenAI SDK to lower the barrier to entry for developers familiar with existing AI tools. It operates on a pay-per-use model, with payments and credits managed through the platform, and is integrated with the Moonbase rewards layer to distribute revenue back to model contributors. ^[7]

Ecosystem Participants

The FLock ecosystem is designed to support various roles, each contributing to the AI model lifecycle. Participation in most key roles requires staking FLOCK tokens to ensure commitment and align incentives with the network's health. ^[2]

• Task Creators: These are individuals or organizations that define AI training objectives and submit them as tasks to the platform. They are required to stake FLOCK tokens or have a proven reputation to create tasks.
• Training Nodes: Participants in the AI Arena who use their computational resources to train models based on the specifications provided by Task Creators. They must stake tokens to participate and earn rewards based on their performance.
• Validators: These participants are responsible for evaluating the models submitted by Training Nodes in the AI Arena. Their role is to ensure the quality, accuracy, and reliability of the models before they are advanced or rewarded. Validators also stake tokens to participate.
• FL Clients: Participants in the FL Alliance who use their private, local data to fine-tune models. They act as either proposers, who train the model with their data, or voters, who validate the proposed updates.
• Model Hosts: Once a model is fully trained and validated, Model Hosts deploy and maintain it on the AI Marketplace, providing inference services to end-users and applications.
• Delegators: Token holders who may not have the technical capacity to act as a node or validator can delegate their stake to active participants. In return, they earn a share of the rewards generated by the participant they support.

These roles create a comprehensive and self-sustaining ecosystem for decentralized AI development. ^[4]

Tokenomics

The FLOCK token is the native utility and governance token of the FLock.io ecosystem. It is designed to facilitate economic activity, secure the network, and enable community-led governance. ^[2]

FLOCK Token

• Ticker: FLOCK
• Total Supply: 1,000,000,000 FLOCK (fixed)
• Blockchain: Base
• Contract Address: 0x5aB3D4c385B400F3aBB49e80DE2fAF6a88A7B691

The token's supply is fixed, and new tokens are introduced into circulation through a scheduled emission process that rewards active and honest network participants. ^[8]

Utility

The FLOCK token has several key functions within the platform that are designed to drive demand and ensure its integral role in the ecosystem:

• Staking: All active participants, including Task Creators, Training Nodes, and Validators, are required to stake FLOCK tokens as collateral. This mechanism secures the network by ensuring that participants have a financial stake in behaving honestly.
• Payments: End-users and developers use FLOCK tokens to pay for access to AI models hosted on the AI Marketplace, particularly for usage beyond any free tiers determined by their stake.
• Governance: Holding FLOCK grants voting rights in the FLock DAO. Token holders can propose and vote on protocol upgrades, treasury management, and other key decisions affecting the platform's future.
• Bounties: Task Creators can offer additional FLOCK tokens as bounties to incentivize participants to prioritize and complete their specific AI training tasks.
• Delegation: The token allows holders to delegate their stake to network operators, enabling them to participate in the ecosystem's reward system without running technical infrastructure themselves.

Economic Model

The platform's economic model is centered around a system of rewards and penalties. Daily token emissions are distributed to contributing participants, with the allocation split between AI Arena and FL Alliance tasks based on the total amount of FLOCK staked in each category. To penalize malicious behavior, the platform employs a "slashing" mechanism, where a portion of a dishonest participant's staked tokens is confiscated and redistributed to honest actors. This creates a strong economic disincentive against attempts to compromise the network. ^[2]

Use Cases

FLock's decentralized and privacy-preserving framework is applicable across various AI domains, particularly those involving sensitive data or requiring community collaboration.

• Large Language Models (LLMs): The platform can be used for both the pre-training of LLMs, leveraging decentralized compute and diverse datasets, and the fine-tuning of existing models for specialized tasks. One such application is creating AI agents for on-chain crypto transactions, with potential hosting on networks like Morpheus. ^[2]
• Stable Diffusion Models: FLock enables community-driven fine-tuning of text-to-image models. Using techniques like LoRA, communities can collaboratively incorporate unique artistic styles and preferences into a shared model without needing to centralize their individual datasets. ^[1]
• Healthcare: The platform's privacy-preserving nature makes it suitable for applications in sensitive fields like healthcare. For example, multiple hospitals could collaborate to train a diabetes prediction model using their respective patient data. Through federated learning, this can be achieved without sharing the raw patient data, ensuring compliance with regulations such as HIPAA. A pilot project in collaboration with a British hospital successfully demonstrated this capability for blood glucose prediction. ^[4]

Governance

Governance of the FLock.io platform is conducted through a Decentralized Autonomous Organization (DAO), which empowers FLOCK token holders to collectively manage the ecosystem. This structure is designed to ensure that the platform's development and policies align with the interests of its community. ^[2]

Proposals can be submitted by any token holder who stakes a specified amount of FLOCK, a measure intended to prevent spam. The community then reviews the proposal before it proceeds to a formal vote. For a vote to be valid, a minimum quorum of token holders must participate. The scope of the DAO's authority includes technical protocol upgrades, treasury management, community initiatives, and adjustments to economic parameters like reward distribution. A key function of the DAO is task verification, where it can approve certain AI training tasks, making them eligible for protocol-level emission rewards. ^[4]

Partnerships and Investors

FLock.io is supported by a number of venture capital firms and has established collaborations with various projects in the Web3 and AI sectors.

Investors

The project has received backing from several prominent investors, including:

• Digital Currency Group (DCG)
• Lightspeed Faction
• OKX Ventures
• Volt Capital
• Tagus Capital
• Ethereum Foundation

Collaborations

FLock has formed strategic partnerships to expand its ecosystem and provide resources to its users:

• Animoca Brands: A task creator on the platform, using FLock to train AI models for enhancing inter-company efficiency within the GameFi ecosystem.
• Base: The blockchain ecosystem where FLock is built. Base has endorsed FLock for its work in creating decentralized, privacy-first AI training infrastructure.
• Morpheus: FLock contributes to the Morpheus Retrieval-Augmented Generation (RAG) Agent for on-chain transactions.
• IO.net: Partnered with FLock on a "Proof of AI" initiative aimed at making decentralized networks a more trusted and scalable solution for AI applications.
• Akash: Enables developers to deploy FLock Training Nodes and Validators on its decentralized cloud computing marketplace.
• Hyperbolic: Provides compute resources to the FLock network to support AI model training.
• United Nations Development Programme (UNDP): FLock was selected as an AI Strategic Partner for the UNDP's Sustainable Development Goals (SDG) Blockchain Accelerator. In this role, FLock provides mentorship and technical support to pilot projects, helping them integrate federated learning and decentralized AI to address global development challenges. ^[8]

The project has also engaged in a flagship collaboration with a leading British hospital affiliated with a UK university to apply its technology to medical AI, as well as partnerships with unnamed leaders in decentralized compute and peer-to-peer smart agent networks. ^{[1] [4]}