Building a Cross-AI Large Model "Privacy Memory Layer", ZetaChain Collaborates with Multi-Model Aggregation Application Anuma to Create a New AI Experience

Author: Zen, PANews

In today’s generative AI applications, users often face fragmented conversational experiences. When switching between different models, the context of previous conversations is often not preserved, requiring users to repeatedly provide the same information from scratch. For example, details discussed on ChatGPT about a certain project cannot be directly inherited when switching to Claude or other models, severely impacting efficiency.

Moreover, the dialogue data of these large models are usually stored on servers of various platforms, leaving users without privacy guarantees and lacking control over their own data. “These real-world issues not only cause a disjointed user experience but also raise concerns about user data sovereignty and security.”

To address this pain point, the industry has begun exploring the concept of a “migratable, user-controlled memory layer,” and blockchain technology may be the key to realizing this goal.

Based on blockchain’s open interoperability, it might be possible to create a privacy memory layer that stores AI context as digital assets, enabling seamless transfer across multiple AI platforms without worrying about “forgetting” past interactions each time a tool is changed, while also ensuring data privacy and sovereignty.

ZetaChain 2.0 Launches, Building a Universal Layer for AI and Web3

In response to the above needs, ZetaChain, a public chain project focused on cross-chain interoperability, has seized the opportunity of AI and Web3 integration. In a roadmap review at the end of 2025, ZetaChain announced plans for “2.0” version, which will introduce new features aimed at the AI era on top of its existing universal cross-chain architecture.

On January 27, 2026, ZetaChain 2.0 officially launched, together with its first AI product—Anuma, a privacy-centric large model aggregation application. According to official sources, ZetaChain 2.0 revolves around three core capabilities:

Private Memory Layer is a protocol-level memory system designed specifically for AI interactions. It aims to bridge the context gap between AI tools, making users’ digital memories truly controllable assets. Based on the privacy memory layer, all user conversations are encrypted and stored, with only the user holding the decryption key; the platform itself has no access. Valuable information generated at different times and models is also controlled by the user, allowing continuous accumulation and transfer to new conversations without being monopolized by any single AI service.

AI Portal is a unified routing and execution layer that enables applications to access multiple AI model providers without being locked into a single platform, supporting availability, fallback, and cost/performance optimization. The AI Portal handles underlying model routing and context bridging, allowing users to freely choose between ChatGPT, Anthropic Claude, Google Gemini, and others for responses, with prior conversation memories supported by the privacy memory layer.

Beyond the protocol itself, ZetaChain 2.0 also provides key capabilities as developer tools (SDK). Developers can directly integrate privacy persistent memory, cross-model switching, and monetization components into their products. This SDK enables applications or AI agents to maintain continuous context across different models and invoke various model capabilities on demand, significantly reducing the cost and complexity of building infrastructure internally.

Mechanically, these three core modules complement each other. The private memory layer offers privacy-first user memory and data support; the AI portal provides ongoing interaction capabilities across mainstream large models; and the SDK ecosystem facilitates efficient third-party development and expansion. This also allows ZetaChain to evolve from a foundational cross-chain protocol into a universal platform serving both Web3 and AI.

Privacy and User Sovereignty at the Core: Anuma Launches and Opens Applications

With the official release of ZetaChain 2.0, the project team also unveiled its first consumer-grade AI product—Anuma. Currently in private beta, Anuma is gradually opening access via an invitation waitlist, allowing users to apply for early access through a public waitlist.

As a large model aggregation application, Anuma integrates multiple mainstream large models, allowing users to invoke different AI engines within a single conversation. It offers the convenience similar to aggregation tools like Poe, supporting OpenAI’s GPT series and Anthropic’s Claude, among others.

When users ask questions, they can specify or switch the model used for responses. Switching between engines is just a matter of clicking, without migrating to another app. Users can choose the most suitable model for their questions within Anuma, and the entire conversation continues seamlessly in the same window.

Technically, thanks to ZetaChain’s privacy memory layer, each conversation in Anuma is encrypted and stored as personal memory, which can be seamlessly migrated to new models or sessions. When starting a new conversation or switching AI models within an existing one, Anuma securely injects relevant context into the target model, enabling it to understand prior background and user intent. This greatly improves cross-model collaboration efficiency by eliminating the need for users to repeatedly explain the same background information.

Traditional Web2 companies leverage their centralized advantages to misuse user data, which has long been deeply resented. Phenomena like platform favoritism and data selling are rampant. This has led to user wariness and concerns about centralized platforms, which extend into the rapidly growing AI field.

Anuma places great emphasis on the privacy of conversation content and user sovereignty. The entire platform employs end-to-end encryption to protect user data. From the moment users input messages on the frontend, the content is encrypted with the user’s key and transmitted to the privacy memory layer for storage. When context needs to be provided to an AI model, it is decrypted either on the user side or in a trusted execution environment before being sent to the model. Throughout the process, conversation records are always kept encrypted on-chain or in transit, preventing even ZetaChain nodes or servers from viewing the content.

This sharply contrasts with traditional AI chat services, where chat logs are often stored in plaintext on servers, risking access or leaks by operators. Anuma, through blockchain and encryption technology, achieves a security level similar to Web3 wallet private key management, making user data only readable by the user. This provides a safer option for sensitive AI applications in fields like legal and medical scenarios, encouraging users to share more private content.

In fact, before Anuma’s launch, some multi-model AI dialogue products had already appeared, such as Poe from the U.S. version of Quora and open-source communities like TypingMind.

Compared to these platforms’ cloud service models and local deployment, Anuma’s on-chain encrypted storage balances privacy and sovereignty. In terms of usability and model richness, Anuma eliminates the cumbersome configuration process of TypingMind, offering a convenient multi-model dialogue experience similar to Poe.

The Backstory of Entering AI: ZetaChain’s Technical Logic and Natural Evolution

The ZetaChain team chose to launch version 2.0 and Anuma at this time because of deep technical accumulation and a clear evolutionary logic.

As the first universal L1 blockchain project, ZetaChain has focused since its 2021 launch on solving blockchain fragmentation issues, aiming to build a foundational network connecting all public chains. Built on Cosmos SDK, it naturally supports interoperability with Ethereum, Bitcoin, Cosmos, and other heterogeneous chains.

Through innovations like CAF, ZetaChain simplifies traditional cross-bridge and wrapping operations into a single smart contract call on one chain, providing unified liquidity and user experience. By the end of 2025, ZetaChain’s mainnet had integrated ten major blockchain networks, including Bitcoin, serving tens of millions of users, with over 225 million on-chain transactions.

On the ecosystem and capital side, ZetaChain has also gained broad recognition. Public data shows the project raised $27 million from investors including Blockchain.com, Jane Street, Sky9 Capital, and others. In 2024-2025, global tech giants like Google Cloud, Deutsche Telekom, and Alibaba Cloud joined as validation nodes, endorsing its security and compliance.

Entering the second half of 2025, with the explosion of generative AI, the ZetaChain team realized that the multi-chain ecosystem and multi-model AI share similar pain points: fragmentation across multiple platforms and systems, requiring a universal layer for integration. They proposed the “AI Universal Platform” strategy, introducing blockchain’s trusted computing and storage into AI, creating blockchain infrastructure for the AI era.

ZetaChain 2.0 embodies this vision. It retains and enhances existing cross-chain functions while adding AI privacy memory and interaction capabilities. This aligns with ZetaChain’s consistent vision: making Web3 friendly to both humans and AI. Moving from a “general-purpose blockchain” to an “AI universal platform” is a natural evolution driven by technological convergence and project mission.

“ZetaChain has achieved large-scale unification at the blockchain experience layer,” said core contributor Ankur Nandwani. “ZetaChain 2.0 extends this approach to AI, enabling next-generation applications and agents to operate across models and blockchains, with default privacy, permissioned memory, and global monetization channels.”

New Paradigm of Deep Blockchain and AI Integration, What Is the Outlook?

The launch of ZetaChain 2.0 and its debut product Anuma mark an important attempt at deep integration of blockchain and AI. Under this system, we see a new paradigm for multi-model AI applications: privacy-first, user-controlled, and cross-platform circulation.

Of course, it should be objectively noted that Anuma is still in very early Private Beta, and the ecosystem is in its initial development stage. Many features and details await feedback from testers, such as support for more models, memory layer capacity and performance optimization, and richer third-party developer tools. This means that in the short term, Anuma cannot replace mature single-platform experiences, and some users will need time to adapt to this new interaction mode.

But the direction represented by Anuma is groundbreaking. In the multi-model aggregation track, Anuma offers a different approach from large corporate solutions. Instead of centralized platforms monopolizing data and model invocation rights, it returns choice and memory to users, achieving trust-minimized coordination through blockchain technology.

As Anuma opens for public testing and features evolve, more innovative applications may emerge on this platform, such as privacy-preserving AI advisors, cross-domain intelligent search assistants, and more. How far this privacy-first multi-model experience trend can go remains to be seen.