Solana Foundation Launches Confidential Compute Beta With TryNitoAI

The Solana Foundation and private AI startup TryNitoAI, known as Nito, launched Solana Confidential Compute on November 13, 2026. The open-source toolkit lets developers verify hardware attestations from Trusted Execution Environments directly on-chain. It currently runs as an unaudited beta on Solana's devnet, meaning no real funds are involved at this stage.

The release consists of two components: an open-source SDK and an on-chain verification program. Together they allow developers to check hardware attestations from TEEs on Solana itself, rather than off-chain where the process normally sits outside public view.

A TEE is an isolated section of a processor where code runs shielded from the rest of the machine. An attestation is the hardware's signed statement about what code is running inside that isolated section. Moving verification on-chain means anyone can inspect whether a given piece of hardware passed the check.

The system supports two attestation formats today: Intel's TDX and AMD's SEV-SNP. NVIDIA GPU attestation support remains in development, which limits coverage for the compute-heavy workloads most AI applications require.

The tools are free and open-source. Any developer can participate in verifying attestation records on the network instead of relying on a single gatekeeper.

Who is building on it first?

Nito is the first platform built on the new infrastructure. The company runs open-weight AI models inside attested hardware, enabling end-to-end encrypted interactions with those models. Its private AI chat is live at trynito.ai, and the company has opened a waitlist for a confidential assistant still under development. The Nito team consists of engineers formerly associated with Switchboard, the oracle network.

The Solana Foundation lists four target use cases for the technology:

• Private AI applications
• Personal assistants
• Agents handling sensitive business information
• Private trading strategies

What does the launch change?

The core proposition is replacing trust with proof. An application built on these tools can point to an on-chain attestation showing what hardware and code it uses, rather than asking users to accept assurances about data handling on faith.

The caveats are substantial. The software is unaudited. It runs only on devnet. No date exists for either the security audit or a mainnet release, and hardware coverage is incomplete until NVIDIA support ships.

There is also a structural limitation worth noting. TEE-based privacy ultimately rests on trusting the chip makers whose hardware issues the attestations. The blockchain verifies the receipt, but it still depends on the hardware that produced it. In plain words: the chain confirms the signature, not the silicon.

How does this fit Solana's broader AI push?

Confidential Compute forms part of Solana's stated ambition to serve as a base layer for AI applications. That effort includes a dedicated solana.ai hub and a series of privacy-focused releases.

The most recent preceded this launch by three days. Solana Privacy Rings went live on mainnet on November 10, 2026, marking a faster path to production for cryptographic privacy than the attestation toolset has taken so far.

The gap is instructive. Privacy Rings reached mainnet; Confidential Compute remains on a test network pending an audit with no schedule attached. In plain words: the foundation shipped the cryptography first and the hardware trust layer second.

Developers can build against the SDK on devnet now, at no cost. Whether the toolset reaches mainnet, and with which hardware vendors verified, depends on an audit that has no published timeline.