More and more blockchains are actively introducing AI capabilities, but the so-called "on-chain AI" also faces many challenges in the development process, the most prominent of which is the cross-chain island problem. This fragmentation not only hinders the full realization of the potential of decentralized AI (DeAI), but also makes it difficult for the industry to continue to have a convincing narrative. If there is no practical scenario that can be widely implemented, how far can DeAI, an over-hyped track, go?
Take blockchain-native autonomous AI Agents (artificial intelligence agents) as an example. Although there is currently a lack of accurate statistics on AI Agents on the chain, the existing data shows that this is a rapidly growing and highly active field. Conservative estimates range from hundreds to thousands of AI Agents deployed on different projects and platforms. However, these Agents are distributed like islands on various chains, similar to the inability of computers to communicate with each other before the popularization of the World Wide Web, making it difficult to unleash their potential.
If the problem with centralized AI is that data is monopolized by large companies, forming data islands, then DeAI has the potential to create new islands at the blockchain level. If interoperability between chains is not prioritized, the potential of DeAI will not be fully realized.
This fragmentation is not just about data being stored in different ledgers. It is also reflected in each chain's unique protocols, smart contract languages, virtual machine environments, consensus mechanisms, and overall operating logic.
For example, a DeAI application built on Ethereum and its EVM may not be able to natively call an AI model deployed on a non-EVM chain (such as Solana) unless a complex and potentially unsafe cross-chain bridging solution is used. Similarly, an AI Agent trained on one chain may have difficulty running properly in the environment of another chain. This results in databases and tools on different chains being isolated from each other, ultimately forming isolated islands of DeAI activity.
This fragmentation problem has also appeared in areas such as decentralized identity systems and medical electronic health records. The root cause is the lack of compatibility between platforms, which ultimately limits the scale of development and application impact.
Building a cross-chain super application (Super AI APP)
The vision of the DeAI community is no longer limited to single-point applications on a certain chain. More and more projects are promoting the construction of "super AI applications".
You can think of it as a highly inclusive platform or a highly integrated service network that can realize a complex combination of AI functions across multiple blockchains, including advanced data analysis, distributed model training, autonomous agent deployment, complex decision-making collaboration, etc. Such a system will not be limited by the resources or limitations of a single chain.
On the one hand, dedicated Layer 1 blockchains such as Bittensor and Gensyn are serving the specific needs of DeAI from the underlying architecture design, such as high-intensity data processing, large-scale computing tasks, or providing incentive mechanisms for AI model operation. This is because many general-purpose L1 chains may not be suitable for dealing with the unique challenges of DeAI.
On the other hand, many well-known DeAI protocols and applications, such as Ocean Protocol and SingularityNET, were originally built on general L1 such as Ethereum, and are now beginning to move towards a multi-chain layout.
As a result, a key debate has surfaced: should we choose to build a dedicated L1 to obtain more adaptable performance but limit the scale of the ecosystem, or build on the existing L1/L2 to gain a wider user reach while accepting the compromise of limited AI features?
Regardless of the path, in order to truly break down silos and expand markets and data sources, successful DeAI platforms in the future will inevitably rely on reliable and available cross-chain capabilities.
Challenges of On-Chain AI
The process of building a super AI application is not an easy one. Here are the three core challenges we are facing:
Technical barriers
Protocol incompatibility
Data fragmentation
Security vulnerabilities of cross-chain bridges
Communication efficiency and transaction speed bottleneck
Data governance and standardization challenges
Establish a unified and effective data governance mechanism in multiple autonomous blockchain networks
Interoperability issues between AI models and agents
The differences in the operating environments of each chain make the migration and collaboration of models and agents complex.
On-chain AI: A long road ahead but a bright future
Despite the numerous challenges, industry players have begun to actively explore standardization and cross-chain solutions. Some leading platforms such as BSC and Solana are also promoting the interoperability of the DeAI ecosystem, although it is still in its infancy.
At the same time, protocols, platforms, and conceptual frameworks are also constantly evolving to build a more interconnected DeAI ecosystem, ultimately achieving true “on-chain AI” that even ordinary netizens can easily use.
This trend is almost irreversible, as the synergy between AI and blockchain brings huge potential. Blockchain can solve the most critical challenges for AI, such as data transparency, trust and incentive mechanisms; while AI can give decentralized systems more intelligent capabilities, such as network optimization, resource scheduling, automatic security auditing, etc.
Author: Dr. Max Li, Founder of OORT and Professor of Columbia University
Originally published in Forbes: https://www.forbes.com/sites/digital-assets/2025/06/12/what-good-is-ai-on-blockchain-if-no-one-can-use-it-easily-in-practice/
