What is FLock?

5 min readFeb 28, 2024


FLock is an end-to-end AI co-creation stack integrating decentralised machine learning on-chain. This innovative platform redefines AI model training, finetuning and inferencing, empowering the general public to contribute knowledge and enrich AI models while preserving privacy.


FLock’s vision is to democratise AI by bridging three foundational pillars: community-owned models, privacy-preserved data, and decentralised computing. The platform fosters diverse AI models, from intelligent agents to sophisticated trading and confluence bots, pioneering a new era of accessible AI tools.

Key Features

  • Community-Owned Models: FLock’s AI models are built by the many, not just the few. The platform enables the community to actively participate in expanding and refining knowledge bases for AI models, ensuring a diverse and rich input source while maintaining strict data privacy standards.
  • Privacy-Preserved Data: FLock employs cutting-edge technologies such as federated learning, zkFL, homomorphic encryption, and Secure Multi-Party Computation (SMPC) to protect user data privacy. Data remains safely on the user’s device, preventing central data collection and potential misuse.
  • Decentralised Compute: FLock leverages decentralised computing power to train and fine-tune AI models. This approach enables scalability, reduces costs, and enhances security.

Our platform is unique in that it empowers the general public to contribute their knowledge, enriching AI models with their data while still preserving privacy, thanks to the fundamental research domain that we created in combination with federated learning and blockchain (check out our papers here: https://scholar.google.com/citations?user=s0eOtD8AAAAJ&hl=en). This community-driven approach ensures that our AI solutions are built by the many, not just the few. Furthermore, we are revolutionising the AI landscape by incentivising AI model training/inferencing/finetuning. This will significantly leverage decentralised computing power, enabling FLock to leverage distributed resources for scalability.

FLock’s ecosystem comprises three participants: trainer, validator and model host, which will bring a new level of decentralisation, safety, and efficiency to machine learning.

In the future, FLock will be the provider of ML models in Web3 and continue to improve them through user contributions and feedback. It is modular and can be plugged into any decentralised hosting network (io.net, Gensyn, Ritual, etc.) to make its models more accessible and to any Dapps to improve their performance.

Products and Use Cases

  1. Co-creation Platform: Our co-creation platform enables the community to actively participate in expanding and refining knowledge bases for AI models. A prime example is BTC-GPT, which achieved 5,000 calls in just three weeks. Allowing the community to contribute, we ensure a diverse and rich input source while maintaining strict data privacy standards. https://beta.flock.io/
  2. Federated Learning Client: We provide a Federated Learning client for developers, enabling them to train and validate models decentralised. This approach to model training upholds our commitment to data privacy by allowing data to remain on the user’s device, preventing central data collection and potential misuse. This client is part of our broader initiative to democratise AI development, making it accessible, transparent, and secure for developers around the globe. This innovative technology has been recognised for its excellence, winning an award at the prestigious NeurIPS conference and highlighting our leadership in creating impactful, privacy-preserving AI solutions. https://github.com/FLock-io/client-interface
  3. A series of use cases already serve users:

a. AI-enabled FVM: Leveraging machine learning to enhance file management and operations.

b. Crowdsourced 0–1 Code Auditing: Utilizing the power of the crowd to improve code quality and security, emphasising our belief in community-driven development.

c. Decentralised Health Tracking & Alerts: Offering a privacy-centric health monitoring solution reflects our commitment to secure and user-focused applications.

d. DID & Credit Score: Pioneering in decentralised identity and credit scoring, illustrating our dedication to revolutionising personal finance through Web3 technologies.

If you’d like to get involved early with our product, please visit our Beta at: https://beta.flock.io/points. Users can accumulate points there.

Are you curious about Flock’s role in the AI ecosystem?

The AI ecosystem is transitioning from Web2’s centralised services (AWS) to Web3’s decentralised networks (such as Render Network and I/O Net), representing a significant leap towards a more inclusive computing ecosystem.

Initiatives like Bittensor have demonstrated the practicality of decentralised resource scheduling. Centralised platforms (such as OpenAI) have streamlined the data processing and model training processes, and now, with FLock technology, communities are incentivised to contribute their idle resources with privacy assurance.

We categorise the Web2 AI and Web3 AI ecosystems into different layers, with FLock.io positioned at the foundational layer of the Web3 AI ecosystem. It fosters decentralised data contribution by incentivising it, coordinates computing power within a decentralised framework for model training and fine-tuning, supports on-chain federated learning, and provides solutions for multiple stages of Web3 AI, thereby enhancing the entire ecosystem.

For an overview of the Web3 AI ecosystem, revisit our previous article: https://twitter.com/flock_io/status/1753027782286676049

What problem are we solving?

FLock is addressing the issue of centralised control over AI models, which leads to restricted access, model bias, and inadequate compensation for data contributors. Closed-source large model providers like OpenAI can monitor all user interactions with the model, including user privacy.

Additionally, since centralised institutions entirely control the ChatGPT model, there is no governance over the model’s output. In many recent cases, we see dramatic amplification of model biases and inaccuracies.

Furthermore, every user of the centralised large language model is essentially an accessible data contributor to such large corporations who own the model. The fairness of contribution incentives and data value assessments must be improved.

FLock has implemented decentralised learning for models, where local models are trained on decentralised training nodes and synchronised with the global model. The nature of Federated Learning ensures user data is kept safely locally. In addition, FLock also employs technologies like zkFL, homomorphic encryption, and Secure Multi-Party Computation (SMPC) to provide extra protection for data privacy. The governance of the whole training process is achieved by blockchain, which is automated, transparent, and immutable.

Additionally, FLock supports various model structures, from complex neural networks to simpler statistical models. Through these aspects, FLock enables decentralised machine learning, thus addressing a series of issues caused by centralised control over AI models.

Reach out to us by

Website: https://flock.io/

Twitter: https://twitter.com/flock_io

Telegram: https://t.me/flock_io_community

Discord: https://discord.gg/flock-io