This article is a submission and does not represent the views of ChainCatcher, nor does it constitute investment advice.

We are pleased to announce that Zypher Network has completed a $7 million financing round to advance our development of decentralized trust infrastructure for AI agents. This round was co-led by UOB Risk Management and Signum Capital, with participation from several institutions including HashKey Capital, Hong Leong Group, Cogitent Ventures, Catcher VC, Hydrogenesis Labs, and DWF Ventures.

Autonomous agents have arrived, but can we trust them?

From OpenAI’s Operator to Nvidia’s Eureka, we are witnessing a historic transformation in the software domain: autonomous agents capable of independent thought, reasoning, and action. These systems are being widely deployed in finance, supply chains, legal services, and infrastructure. By 2035, agent systems are expected to drive an economic scale of $250 billion to $350 billion, reshaping over $10 trillion in the “on-demand services” industry.

However, autonomy without accountability is a dangerous proposition. Current agent systems resemble opaque black boxes, with behaviors that cannot be verified and logic that is inaccessible. In sensitive environments, this lack of transparency poses existential risks—from financial decision-making errors to privacy breaches and regulatory non-compliance.

The internet of the 1990s faced a similar crisis. Without encryption or endpoint verification, trust was extremely fragile. Subsequently, SSL emerged as a standardized layer for secure communication. We believe that the current agent systems are approaching their own SSL moment—cryptographic verifiability will become foundational. Zypher Network is building this trust layer for AI.

Our story: From verifiable applications to agentic infrastructure

Zypher originated in 2023, with a founding team spread across Hong Kong and Silicon Valley, initially developing the first verifiable applications based on ZK co-processors (zero-knowledge co-processors), including on-chain reasoning engines, compliance circuits, and award-winning fully on-chain games. These early products attracted over 1 million on-chain participants, showcasing the potential of real-time proof-driven computation.

But we are looking to the future. We asked ourselves: What are the most pressing and impactful applications for the next decade? The answer is clear—AI agents. Systems based on large language models are evolving into autonomous actors, yet there is no cryptographic framework holding them accountable. We have locked our mission onto this, transforming into verifiable AI infrastructure and officially launching Zypher Network, developing a suite of zero-knowledge protocols suitable for agent trust. In 2025, we expanded this vision through a $7 million financing round, gaining support from investors who resonate with our long-term vision.

Challenges: Trust barriers for decentralized AI

The pace of AI development is outstripping its governance frameworks and institutions. Agents are executing workflows in finance, healthcare, legal services, and logistics, yet most systems lack visibility and auditability. Since 2023, AI agents based on large language models have reshaped the computing paradigm, enabling agents to perform complex tasks with minimal supervision.

An industry survey in 2024 revealed that 65% of companies adopting AI agents list trust and security as their top concerns, with the finance and asset management sectors facing the highest risks. Without decentralized solutions, companies rely on centralized systems, creating single points of failure and violating Web3 principles. Zypher guarantees agent behavior through cryptography, providing the following verification capabilities:

• The exact system prompts or instructions received.
• The generated outputs or reasoning results.
• The results have not been modified and are faithfully transmitted.
  This achieves verifiable autonomy—agents can accept accountability from systems and users while maintaining independence.

Our infrastructure: End-to-end verifiability for autonomous AI

Zypher’s architecture combines a modular zero-knowledge protocol stack with a performance rollup optimized for real-time decentralized AI agent verification—Zytron.

● ZKP-based trust solutions
Our open-source ZKP protocol ensures that AI prompts and reasoning are tamper-proof and privacy-preserving. The flagship solution “Prompt Proof” (zkPrompt) is a groundbreaking innovation. Inspired by zkTLS, zkPrompt employs a Prover-Proxy-LLM architecture to verify AI outputs.

• The Prover generates ZK proofs, confirming the integrity of the agent’s response.
• The agent’s signature mechanism supports on-chain verification.
  Compared to frameworks like ezkl, zkPrompt reduces proof generation time by up to 40%, making it suitable for real-time applications.
  For example, DeFi protocols can use zkPrompt to prove that an agent’s portfolio management decisions comply with predefined strategies without revealing the strategies. In asset management, zkPrompt ensures compliance while protecting sensitive data.

Our ZKP suite also includes reasoning verification and model integrity protocols, providing a comprehensive trust layer for AI. We embed this into an open-source RESTful API, allowing developers to seamlessly integrate trust features, whether for Web3 native tools or enterprise solutions.

● Zytron Rollup
Zytron is our dedicated rollup infrastructure, serving as a Layer 2 solution for the BNB chain, optimized for AI workloads with high throughput, low latency, and robust security. Compatible with RISC-V architecture, Zytron supports modular integration with various AI frameworks (from large language models to specialized models).

Key components of Zytron include:

• Distributed proof protocol: Based on a “verifiable work proof” mechanism, enabling a decentralized network of nodes (“provers”) to compute ZK proofs, ensuring computational integrity. Unlike energy-intensive proof-of-work, our system emphasizes efficiency and fair reward distribution.
• Computational resource integration: Zytron connects distributed computing resources, supporting AI model hosting, sharing, reasoning, and fine-tuning, giving developers access to scalable infrastructure.
• API and proof layer: Provides an API layer for seamless model access and a proof layer for verifying computations, ensuring privacy and resistance to censorship.

Zytron addresses scalability challenges by processing thousands of proofs simultaneously, avoiding traditional blockchain bottlenecks. For instance, payment systems can use our API to validate AI-driven transactions in near real-time under high demand. Leveraging the security of the BNB chain, Zytron ensures tamper-proof verification, establishing trust for high-risk AI applications.

● More about Prompt Proof: zk-TLS-like protocol
zkPrompt is Zypher’s flagship protocol, inspired by zkTLS, based on a lightweight zk circuit system that proves the integrity of LLM agent outputs. It introduces four roles:

• User: Initiates requests.
• Prover: Processes transmissions and delivers proofs.
• Agent: Relays data and signs ciphertext.
• LLM Provider: Generates agent responses.

To prevent tampering, zkPrompt uses agent-based signature verification and efficient ZK decryption correctness proofs. The prover must demonstrate that the returned plaintext matches the agent’s signature and the original content generated by the LLM provider.
The system supports two modes:

• Public mode: Prompt and response data are on-chain.
• Private mode: Only hash commitments are recorded, protecting privacy.
  Compared to ezkl, zkPrompt is optimized for real-time verification, suitable for minute-level proof generation for GPT-like models, with minimal overhead.

● AI Secure Browser: An SSL moment for everyday users
For the average user, we designed the AI Secure Browser—a user-friendly real-time monitoring tool that brings cryptographic integrity directly to the application layer. Just as the padlock symbol and HTTPS standard established trust in the early internet, the AI Secure Browser makes verifiability visible and accessible to end users.

This browser acts as a security overlay on any AI interface, providing:

• Proof-based trust indicators: Each agent interaction comes with a visual marker indicating whether its output has been cryptographically verified against the original prompt via zkPrompt.
• Anomaly alerts: Detects tampered outputs, unauthorized queries, or unverified completions, and prompts immediately.
• Verifiable interaction logs: Users can explore the history of agent interactions, viewing the time, manner, and conditions under which responses were generated, all supported by zero-knowledge proofs.
• Privacy-first operations: Users have complete control over disclosed information. For sensitive use cases, proof commitments can be published without revealing prompt content.

To make trust participatory, the browser also integrates our proof mining incentive layer:

• Users earn GEM points and collectible NFTs by verifying agent outputs.
• Active participants build trust by flagging anomalies, verifying cryptographic claims, and enhancing decentralized AI security.
  In short, the AI Secure Browser is to autonomous agents what SSL is to the web: a foundation of integrity at the user level. It brings transparency to previously opaque systems, providing real-time assurance to every user, from DAO contributors to legal teams and financial operators.

Combined with our developer stack, the AI Secure Browser completes Zypher’s vision of a full-stack trust layer for verifiable AI, enabling both system creators and consumers to operate securely and confidently.

● Our roadmap: Expanding AI trust
This $7 million financing propels our ambitious roadmap:

• Team expansion: Expanding engineering and research teams in Hong Kong, Silicon Valley, and globally to drive cutting-edge cryptography and AI innovation.
• Infrastructure upgrades: Optimizing Zytron rollup and ZKP protocols to support large-scale AI deployments, improving proof generation efficiency and enhancing model compatibility.
• Developer and ecosystem growth: Launching new incentive programs and tools to empower developers to use our API and zkPrompt suite across DeFi, customer support, and robotics.
• Token release preparation: Preparing for a token release based on strong ecosystem appeal to further incentivize participation and decentralization.
• Deepening protocol integration: Our trust layer has integrated with leading protocols, enhancing cross-chain interoperability and developer experience.
  Our vision is to make Zypher the cornerstone of trusted AI in Web3, enabling developers to create applications that rival centralized systems while upholding decentralized principles.

● Our community: The heart of Zypher
Our community is our greatest asset. Currently, over 1 million on-chain participants have joined our journey through ecosystem programs, API testing, and collaborative activities.

This momentum is also supported by a growing network of strategic partners. Zypher’s trust layer has integrated with leading protocols such as Eliza OS, Mira Network, io.Net, Nexus, Risc Zero, EigenLayer, Particle Network, Fermah, Polyhedra, and ZeroBase, paving the way for interoperability, developer onboarding, and real-world use cases. We take pride in collaborating with the most forward-thinking protocols and developers in Web3. The community has always been at the core of our mission: to make AI trustworthy, transparent, and verifiable for everyone.

In the coming months, we will launch large-scale social and proof mining activities to incentivize decentralized agent verification. These initiatives have already attracted early participation from key agent providers in the network.
About Zypher Network
Zypher Network is a decentralized trust platform that achieves verifiable autonomy for AI agents through zero-knowledge protocols and dedicated rollup infrastructure Zytron. Zypher operates in Hong Kong and Silicon Valley, empowering developers and enterprises to build secure, scalable AI systems for Web3 native and real-world applications.