A bank that holds your assets answers to vault procedures, dual sign-off, regulators and outside auditors. All of that exists to stop the people with access to your funds from misusing it, and most customers never think about any of it.
Blockchains were supposed to remove the need for that machinery. Mostly, they just moved it. If you want your Bitcoin to interact with an application on another network, someone is holding a key on your behalf, usually a bridge or a custodian. The history of this industry is a long list of the times that went wrong.
This week, SVRN announced the completion of work we’ve led since January on NEAR Chain Signatures, the system that lets NEAR accounts and AI agents transact on more than 30 blockchain networks. We designed the program that funds its 17 independent operators, took it through governance, vetted the operators that joined, and ran one of the nodes ourselves. Earlier this week, the last of those 17 nodes finished moving into hardware whose contents anyone can verify. For anyone who holds NEAR, or holds SVRN, I’d put this among the more consequential things the network has shipped this year, and it’s easy to miss because all of it happens below the application layer.
What are NEAR Chain Signatures?
Chain Signatures let a single NEAR account hold and move assets on Bitcoin, Ethereum, Solana and more than 30 other networks without a separate wallet and private key for each one. The closest comparison in traditional finance is correspondent banking, which lets your local bank move money in currencies it doesn’t hold. The difference is that here no correspondent ever takes custody of your funds.
NEAR gets there by making sure a complete key never exists anywhere. Each key is generated in pieces and spread across 17 independent operators, and a transaction is only signed when at least 11 of them cooperate, so no single operator, SVRN included, can move anything on its own. If a node goes offline, the network routes the request to the others. The technique is called multi-party computation, or MPC, and I wrote about how this kind of threshold signing fits into AI agent custody earlier this month.
This is the layer underneath NEAR Intents, which has processed more than $34 billion in cross-chain volume, and virtually all of the Bitcoin and Solana flow crossing into NEAR now moves through it. Near One publishes a live map of the network where you can watch signing requests from the Bitcoin and Zcash transfers arrive every few seconds, along with which operator assembled each signature.
What changed when the network moved into Intel TDX?
For its first two years the network’s security rested on a reasonable assumption: that 17 independent operators, running on ordinary servers, would not be compromised at the same time. The assumption held up, but anyone evaluating the system still had to take each operator’s word for the state of its machines.
Over the past few weeks every operator moved its key share into a trusted execution environment (TEE) built on Intel TDX, a sealed hardware environment that the operator’s own administrators, its hosting provider and the server’s operating system can’t read. I can’t read SVRN’s own key share, and I consider that a feature. The hardware also produces a cryptographic proof of exactly what software is running inside it, which anyone can check against the network’s published code, and any upgrade to that software now has to be voted in onchain by the operators, where anyone can rebuild it and confirm what was approved.
Anyone evaluating the system can now check the machine instead of taking the operator’s word that its servers are clean, and that is the difference an auditor cares about. Every core component has been audited by Trail of Bits, and the reports are public.” Hardware isolation only protects you when attestation is actually enforced, which David Schwed covers in TEEs aren’t a checkbox.
The same isolation keeps the signing process out of view of the people running it, which is the basis for confidential transactions on NEAR, where the infrastructure carrying a transaction never sees its details.
Why do AI agents need verifiable signing?
The next wave of transactions will be initiated by AI agents acting for people and businesses: paying for compute, settling invoices, rebalancing positions on standing instructions. Those agents run around the clock, across networks that don’t share a security model, and they carry information their principals have every reason to keep private.
If the keys sit with an intermediary who can also read every instruction, that intermediary becomes the most valuable target in the agent economy, and targets like that tend to get hit. NEAR’s signing layer is now built so that nobody, the operators included, is in that position. Agent payments still need other infrastructure, and the rest of it sits on this layer.
Why would a company that manages a NEAR treasury help build the network?
SVRN’s treasury holds more than 55 million NEAR, and our commercial strategy depends on enterprises trusting this network with real transactions. Few companies have as much riding on the integrity of NEAR’s infrastructure as we do.
A treasury that only holds the token is a leveraged bet on someone else’s execution, and helping run the network was always part of how we intended to participate in NEAR. In January, Kendall Cole of Proximity Labs and I wrote HSP-007, the proposal that replaced discretionary grants for these operators with a standing program. We built support for it across NEAR’s stakeholders and took it to a vote, and the House of Stake, NEAR’s governance body, approved it in February. The program funds operators out of protocol emissions through the House of Stake, monthly and in arrears, and only for performance: full payment at 99% uptime or better, half below that, nothing below 95%. Operators can challenge each other’s uptime, and every payment and result is published monthly, so a NEAR holder can see what the network is paying for and whether it got it.
Once the program passed, we led the diligence that qualified each new operator. Every node was assessed against the program’s requirements before it joined the signing set, so anyone holding a piece of the network’s keys had shown it could meet the standard. Then we held ourselves to the same bar, with our own node that went live on mainnet in July. I think of this as the third leg of how SVRN participates in NEAR: we hold the asset, we bring enterprises onto the network, and we help govern and secure the infrastructure both of those depend on.
What does this mean for SVRN shareholders?
If you own SVRN, you have exposure to NEAR’s growth through the treasury. You also own a company that wrote the rules for the infrastructure that growth depends on and now helps run it, paid for its performance and accountable to a public governance process for its conduct. As more cross-chain volume moves through Intents, and more of it is initiated by agents, this signing network is what it runs through, and SVRN is one of the 17 parties running it.
The move into verifiable hardware was a coordinated effort across all 17 operators and the Near One team that built the system, with Anton Astafiev leading it to completion. The program has room for 21 operators, and we’ll keep helping qualify the ones that join. NEAR Foundation’s overview of the program covers how operators are funded, measured and governed, and the live operator page shows the full set and each node’s hardware status.
overview
