SecretVM and Solidity-LLM introduce Confidential AI for Web3

In a breakthrough that combines privacy, cryptographic integrity, and intelligent automation, an AI model specifically designed for smart contract development has been successfully deployed inside confidential virtual machines on the Secret Network. Solidity-LLM, the large-scale language model developed by ChainGPT, runs within SecretVM, the Confidential Virtual Machine framework that powers Secret Network’s latest evolution in distributed computing.

This is the first time that an AI model trained specifically for writing and auditing Solidity code has been incorporated into a trusted execution environment. The impact on blockchain development is significant. Developers can now leverage artificial intelligence to generate, optimize, and analyze smart contracts without exposing their source code or proprietary logic to third parties. Privacy is protected by design and verification is built into the system’s cryptographic structure.

At the heart of this development is SecretVM, a privacy-preserving computing layer built on top of hardware-based TEEs such as Intel’s TDX and AMD’s SEV. SecretVM provides confidentiality, execution integrity, and remote authentication in a single runtime. This allows you to run sensitive workloads such as AI inference, financial calculations, and cross-chain reconciliation within a sealed enclave. In reality, even node operators have no access to the data, inputs, and outputs processed by the AI.

“Secret computing is no longer an abstract concept,” said Luke Bowman, COO of the Secret Network Foundation. “We have shown that we can run complex AI models purpose-built for Solidity within a fully encrypted environment, and that all inferences can be verified on-chain. This is a true milestone for both privacy and decentralized infrastructure.”

Also read: AiThority Interview Featuring: Pranav Nambiar, Senior Vice President, DigitalOcean AI/ML and PaaS

Solidity-LLM was developed by ChainGPT to address the specific needs of Ethereum and EVM compatible smart contract development. It is a 2 billion parameter model trained on over 650,000 carefully selected Solidity contracts. This domain-specific approach allows the model to provide a nuanced understanding of contract logic, optimization patterns, and security practices that cannot be reproduced by general language models. Currently, this model, running inside SecretVM, operates completely privately, enabling confidential on-chain software development for the first time.

“We’ve spent the last year building and fine-tuning a model that understands the actual structure of Solidity,” said Christopher Duggan, Head of Marketing at ChainGPT. “What makes this introduction so important is that developers can now use the model without sacrificing IP or exposing critical business logic. Everything remains encrypted and verifiable.”

This introduction resolves a long-standing paradox in AI-assisted development. Traditional coding assistants require developers to upload their source code to a centralized cloud platform. Doing so risks exposing intellectual property, user data, and business logic. However, within SecretVM, Solidity-LLM performs inference in an encrypted and verifiable environment. Cryptographic proofs ensure the integrity of the enclave. This means developers and institutions can trust not only the output of AI, but also the infrastructure on which it operates.

The architecture that supports this deployment is robust and flexible. Solidity-LLM runs as a containerized workload within SecretVM and leverages Docker to support a variety of language frameworks and integration methods. Access is provided via APIs, SDKs, or direct smart contract interfaces. Developers can treat models as programmable agents in their own right, allowing them to interact with contracts, tools, and governance frameworks without revealing sensitive inputs.

Use cases are wide-ranging. For developers and builders, this means confidential collaboration. You can generate, test, and optimize your code privately without worrying about your proprietary logic being exposed. The system provides auditors and security professionals with a trusted environment for AI-assisted contract review and analysis. Data never leaves the enclave and remains confidential throughout the audit process.

This architecture provides a clear path to blockchain automation for institutions and enterprises without the compliance and privacy risks associated with traditional AI tools. Smart contracts can now be generated and deployed in a manner that aligns with regulatory standards, especially in sensitive sectors such as finance, healthcare, and governance.

The implications for decentralized finance and DAOs are particularly compelling. Smart contract agents can now operate autonomously within a cryptographically validated environment. Manage upgrades, make governance decisions, and orchestrate cross-chain logic without exposing internal state or decision-making data.

Researchers and AI engineers will also benefit. SecretVM supports federated training and fine-tuning across encrypted datasets. This enables collaborative AI development without compromising data sovereignty. This enables secure multiparty computing, where models evolve alongside distributed networks rather than within corporate silos.

This integration changes the AI ​​on-chain trust model. We move from opaque clouds to verifiable computing, from public exposure to private collaboration, and from centralized inference to decentralized autonomy. Developers retain ownership of their data, code, and infrastructure.

The roadmap includes fine-tuning sensitive models, orchestrating multiple AI agents, and cross-chain deployment. As artificial intelligence continues to merge with blockchain infrastructure, the fundamental challenge remains: how to build trust without relinquishing control. This expansion provides practical and proven answers.