
The concept of decentralised artificial intelligence (DeAI) has recently gained significant traction, with experts and institutions debating its feasibility, challenges, and potential impact. Unlike traditional centralised AI models, which are controlled by a few powerful organisations, DeAI aims to distribute AI capabilities across a broad network, ensuring greater accessibility, transparency, and efficiency.
Leading voices in AI research and development, such as the MIT Media Lab, the Linux Foundation, and major media outlets like Forbes, have all weighed in on DeAI. MIT Media Lab emphasises the need for personalised AI agents and the democratisation of AI, countering the current monopolisation of the industry.
It highlights the challenges of centralised AI, such as limited data access due to data silos, lack of transparency, and concerns about trust and accountability. Their call for businesses to adopt decentralised models is echoed by the Linux Foundation, which published a detailed 54-page report in November 2024 outlining how autonomous AI agents can function independently within decentralised networks while maintaining privacy, exemplified by zero-knowledge proofs.
Meanwhile, Forbes published an article in February 2025 underscoring the benefits of open-source AI, advocating against AI models being locked behind paywalls and proprietary systems. Their article stated, “Success in AI relies on collective input, demands vast, diverse datasets, and continuous collaboration.”
The question remains: How can the vision of DeAI be turned into reality? What technical challenges must be overcome to achieve decentralised AI, and what role can AI frameworks play in this transformation?
Challenges and solutions in AI frameworks for DeAI
The foundation of DeAI lies in robust AI frameworks that enable AI agents to operate in a decentralised environment. However, existing frameworks are not yet optimised for this shift. Here’s a list of the key challenges that AI frameworks face along with their solutions.
High technical barrier-to-entry for non-developers and democratising AI agent development
Most AI development frameworks require a deep understanding of programming, machine learning models, and infrastructure deployment. The complexity of developing AI models and deploying them in real-world applications limits participation to a small group of highly skilled engineers and data scientists. This restricts the widespread adoption of DeAI by non-technical users and organisations that could otherwise benefit from AI-driven solutions.
Solution: The democratisation of AI agent development should be a top priority for companies. Democratisation can be achieved by fostering collaboration between AI companies or open-source frameworks that remove high technical barriers-to-entry for non-developers.
An example is aevatar.ai, an open-source no-code framework for AI agents, which allows anyone to create, deploy, and utilise AI agents using an intuitive prompt system. Its first use case is a multi-LLM-driven mining ecosystem called MineAI. Mine AI is the first AI agent PVP mining system, allowing users to utilise natural languages as prompts to mine, defend, and attack with dynamic strategies. By eliminating technical barriers, these platforms expand AI adoption and innovation.
Isolated AI agent ecosystems and multi-LLM and multi-agent AI frameworks
Today’s AI frameworks are restricted by proprietary systems, preventing seamless communication between AI agents across different platforms. These isolated ecosystems hinder collaboration, limit interoperability, and prevent AI agents from leveraging complementary capabilities.
This results in inefficiencies where AI models cannot share insights, making them less effective in tackling complex tasks that require diverse knowledge sources. While the Model Context Protocol (MCP) is emerging as a promising solution to address this challenge, several unresolved pain points persist, requiring further development and refinement to ensure seamless agent-to-agent communication.
Solution: A true DeAI ecosystem requires seamless interaction between multiple AI agents, even if they operate on different language models and platforms. Since different LLMs serve different purposes, creating a multi-LLM AI framework that enables siloed AI agents to communicate with one another seamlessly would allow AI agents to produce more holistic results.
Also Read: Navigating market trends and risks: Leveraging GenAI in banking treasury functions in APAC
Developing a universal translator that supports agent-to-agent communications across different languages, platforms, and industries is crucial to ensuring each agent can adapt to diverse scenarios and optimise performance. A decentralised communication protocol for AI agents would further eliminate reliance on centralised intermediaries, enabling AI to function autonomously across various domains.
Lack of scalability in AI frameworks and enhancing scalability through multi-model AI frameworks
Current AI frameworks often depend on a single language model (LLM), restricting their flexibility. A single LLM limits the adaptability of AI agents, especially in decentralised networks where multiple AI agents must interact across diverse environments and applications.
Additionally, computational constraints limit the scale at which AI models can operate efficiently. As AI systems grow in complexity, bottlenecks in processing power, data throughput, and response time become increasingly apparent, making it difficult to deploy large-scale AI networks efficiently.
Solution: Instead of relying on a single LLM, decentralised AI networks must support multiple models that work together to optimise decision-making. This approach enables AI agents to process data with much higher relevance and adapt to changing contexts.
Scaling AI frameworks through decentralised computing infrastructures, such as distributed AI model hosting, will further enhance scalability and reduce reliance on centralised cloud providers. Additionally, leveraging modular AI architectures—where individual AI components can be dynamically loaded and updated—would enable more efficient scalability and adaptability to evolving tasks and requirements.
Lack of decentralised access to off-chain data and AI oracles as a bridge between AI and blockchain networks
AI agents operating within decentralised frameworks often lack access to off-chain data, which limits their functionality and decision-making capabilities. Without reliable access to external datasets, AI agents risk becoming isolated and ineffective in real-world applications.
Solution: AI oracles act as intermediaries, allowing smart contracts to access and process off-chain data securely and verifiably. AI oracles are crucial in enabling decentralised AI networks to interact with real-world data while maintaining decentralisation. By integrating AI oracles, decentralised AI agents can operate in a dynamic environment without relying on central authority to validate data.
Also Read: AI disruption unveiled: Hidden opportunities for startup survival and success
This ensures AI-driven decision-making remains decentralised while benefiting from external real-time data feeds. Additionally, AI oracles leveraging cryptographic techniques such as zero-knowledge proofs can ensure data authenticity without compromising user privacy, further strengthening trust in decentralised AI ecosystems.
The future of DeAI
By addressing these key challenges and implementing advanced AI frameworks, we can realise the vision of a decentralised AI ecosystem. The future of DeAI would have features such as an open-source marketplace for AI agents. A decentralised AI marketplace would allow individuals and enterprises to discover, develop, manage, and deploy AI agents at scale automatically. This ecosystem would function similarly to open-source software repositories, fostering collaboration and innovation among AI developers and users.
In addition, customisable AI agents for business and personal use should be a fundamental feature. Enterprises and individuals will be able to generate AI agents tailored to their specific needs, integrating them seamlessly into existing workflows. These AI agents could range from customer support bots to highly creative AI-driven content generators.
By enabling a user-friendly, scalable, and transparent DeAI system, we can ensure that AI development and deployment remain accessible to all, rather than being monopolised by a few large corporations.
Conclusion
The shift towards decentralised AI requires overcoming significant technological and structural challenges. Current AI frameworks must evolve to reduce technical barriers, foster interoperability, and enhance scalability. By leveraging AI oracles, multi-LLM frameworks, and no-code AI development platforms, we can move closer to a truly decentralised AI ecosystem.
The future of AI should not be locked behind proprietary walls—it must be open, collaborative, and accessible to all. DeAI is not just a vision; it is an achievable reality if we optimise AI frameworks for a decentralised future.
—
Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.
Join us on Instagram, Facebook, X, and LinkedIn to stay connected.
Image courtesy of the author.
The post Optimising AI frameworks for a decentralised AI (DeAI) future appeared first on e27.