Solana launches private AI tools with on-chain hardware checks
The Solana Foundation released a toolkit that allows developers to build private AI apps while verifying the underlying hardware on the blockchain.
The short version
- The toolkit supports hardware attestations from Intel TDX and AMD SEV-SNP, with NVIDIA GPU support currently in development.
- AI processing occurs off-chain, while the verification records for the hardware and signing keys are stored on the Solana blockchain.
- The system uses Trusted Execution Environments (TEEs) to isolate code and data from the host operating system and its operators.
- The tools are currently in beta on the Solana devnet and are intended for developers to build and test applications before a mainnet release.
Why it matters
This allows businesses and individuals to use AI for sensitive tasks like private trading or confidential business analysis without the AI provider seeing the raw data. It creates a framework where users can verify that the hardware running their data is secure and untampered.
A toolkit for private AI
The Solana Foundation recently introduced Solana Confidential Compute, a new open-source toolkit designed for building private AI applications. Developed in coordination with Nito, the toolkit provides a software development kit and an on-chain verification program. It aims to support applications that handle sensitive information, such as private trading strategies, personal conversations, and confidential business data.
The project is currently in a beta phase on the Solana devnet, with a mainnet release planned for the future. While the tools are currently unaudited, they are intended for developers to use while building and testing their specific applications. Nito is already incorporating this infrastructure into its own private AI platform, but other teams can also use the open-source components.
The Foundation described the release as a way to achieve "Private AI, verified onchain." By providing these tools, the network hopes to enable a new class of applications where privacy is a core requirement. This release follows a broader strategy by Solana to prioritize developer tools that handle privacy and confidential data.
How hardware verification works
The system relies on a concept called confidential computing, which protects information while a machine is actively processing it. It uses Trusted Execution Environments, or TEEs, which are secure areas of a processor that isolate code and data from the rest of the system. This isolation ensures that even the host operating system or the person running the server cannot see what is happening inside the protected environment.
Inside a confidential virtual machine, the software generates a unique signing key that stays within the protected hardware. The system then creates a hardware-signed report that links this key to a specific Solana slot. A verification program then checks this report before recording the key, the hardware type, and the expiration date on the blockchain.
When the AI produces an answer, it signs the output with the protected key. An application receiving that answer can then check the signature and look at the corresponding record on Solana. If the records match, the application can trust that the output came from a verified, secure machine without the intermediary ever seeing the private inputs.
Privacy and data flow
In the provided chat template, the browser verifies the confidential environment before any data is sent. Once verified, the browser encrypts the user's prompt before it leaves the device. This means that the message remains encrypted as it passes through the application's server and the Nito infrastructure to reach the protected environment.
The design ensures that neither the application server nor the Nito infrastructure can read the contents of the messages. However, the documentation notes that certain metadata remains visible. For example, the application server can see a visitor's IP address, while Nito can see the account, the model used, the request timing, and the size of the encrypted message.
Developers must still manage certain aspects of the application themselves. They are responsible for implementing authentication, rate limiting, and bot protection. Additionally, the verification process identifies a published software release but does not guarantee that the specific release is free of bugs.
Future applications and goals
The Solana Foundation highlighted several potential use cases for these tools, including personal assistants and agents that work with confidential business data. They also suggested applications for AI-driven rulings in prediction markets or insurance claims. Other possibilities include trading orders that are only accepted from verified machines and AI-driven governance decisions.
These tools align with previous developer priorities outlined by Solana, such as creating an open AI inference marketplace. In that vision, users could verify which specific model handled a request and track how well the provider performed. Another proposal suggested a wallet-based financial assistant that would understand a user's holdings and previous decisions with explicit approval.
While Solana Confidential Compute focuses on the hardware and verification side, other privacy products are also emerging on the network. For instance, Helius recently launched Solana Privacy Rings for encrypted balances. These different layers of privacy work together to create a more comprehensive ecosystem for confidential data on the blockchain.
What to watch
- The official release date for the mainnet version of the toolkit.
- The development progress of NVIDIA GPU attestation support.
- How many third-party developers adopt the open-source components for their own AI platforms.
Try it
If you are a developer, you can explore the Solana Confidential Compute documentation to see the Next.js and React chat template.
Sources and links
- Officialsolana.ai/blog/nito-solana-confidential-compute solana.ai
- ReportingCrypto News: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by Crypto News · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
