Solana launches private AI tools with on-chain hardware checks

The Solana Foundation has introduced Solana Confidential Compute, an open-source toolkit built with Nito that lets developers create private AI applications and verify their computing hardware on Solana.

Summary

• The toolkit is in beta on Solana devnet, with a mainnet release planned.
• Intel TDX and AMD SEV-SNP attestations are supported; NVIDIA GPU verification remains in development.
• AI processing takes place off-chain, while applications can inspect hardware-verification records on Solana.
• The tools remain unaudited and are currently intended for developers building and testing applications.

The Solana Foundation's Nov. 13 announcement describes a software development kit and an on-chain verification program developed by Nito in coordination with the Foundation. Nito is incorporating the infrastructure into its private AI platform, while other teams can use the same open-source components.

"Private AI, verified onchain," the Foundation wrote on X, introducing tools for applications that handle personal conversations, sensitive business information, and private trading strategies.

The beta supports hardware attestations from Intel TDX and AMD SEV-SNP, with NVIDIA GPU attestation still being developed. In the same thread, the Foundation described the release as unaudited and intended for development. The announcement says a mainnet version is coming, without giving a release date.

Solana Confidential Compute checks the hardware behind AI

According to the Foundation, confidential computing protects information while a machine processes it. The system uses trusted execution environments, or TEEs, which isolate code and data from the surrounding infrastructure, including the host operating system and its operator.

Inside a confidential virtual machine, the software generates a signing key that remains within the protected hardware. A hardware-signed report links the key to a recent Solana slot, and the verification program checks the report before recording the key, hardware type, and expiration date.

When the machine produces an answer, it signs the output with that key. An application receiving the answer can check its signature and consult the corresponding record on Solana before accepting it.

The Foundation says the AI computation happens off-chain. Publishing the verification record does not require publishing a user's conversation or private inputs.

Private chat encrypts prompts before they leave the browser

Solana's chat template provides a developer example using Next.js and React. Its documentation says the browser verifies the confidential environment before encrypting a message to its key. Failed verification prevents the prompt from being sent.

Encrypted messages pass through the application's server and Nito to the protected environment, where the model reads the prompt and generates a response. The reply is encrypted before returning to the browser. Neither intermediary can read the message contents under the documented design.

Privacy has specific limits. The template documentation says the application server can see a visitor's IP address, while Nito can see the account, model, request timing, and encrypted message size.

Developers still rely on the webpage code they serve, the hardware manufacturer's attestation, and the published software running inside the environment. Verification identifies a published software release; it does not establish that the release contains no bugs.

The example requires a funded Nito account and API key. Authentication, rate limiting, and bot protection are left for developers to implement.

Developers can build agents around verified outputs

In its X thread, the Foundation described personal assistants, agents working with confidential business data, and private trading strategies as potential applications. It said developers could check a signing key's attestation status before accepting signed results.

The launch announcement lists further possibilities, including AI rulings for prediction markets or insurance claims, trading orders accepted only from verified machines, and AI governance decisions.

Solana had outlined related developer priorities on Nov. 9. As crypto.news previously reported, its 11 startup ideas included an open AI inference marketplace where users could verify which model handled a request and track provider performance.

A separate proposal called for a wallet-based financial assistant that understands holdings and previous decisions, with explicit user approval before executing actions.

Solana privacy products cover payments and token allocations

Helius launched Solana Privacy Rings on Nov. 10, offering encrypted balances and confidential transfers on mainnet. The initial permissionless ring hides assets and amounts while leaving sender and recipient addresses visible.

Earlier, Umbra and Streamflow introduced confidential token vesting on May 26. The integration encrypts allocation amounts, recipient addresses, and unlock schedules using Arcium's encrypted computing infrastructure.

In the United States, SEC Chairman Paul Atkins discussed financial privacy in December 2025 remarks. Speaking in a personal capacity, he said zero-knowledge proofs and selective disclosure could help users demonstrate compliance without revealing their entire financial history.