Blockstream, the Bitcoin infrastructure company co-founded by cryptographer Adam Back, has rolled out a new institutional custody offering designed to strengthen offline-style security while still supporting practical transaction approvals. The service, announced Friday in a collaboration with French technology group Thales, pairs Thales’ hardware security modules (HSMs) with Blockstream’s Secure Multi-Party Computation (Secure-MPC) technology.

According to Blockstream, the solution is built so that institutions can authorize Bitcoin transactions while keeping the most sensitive cryptographic steps confined inside dedicated security hardware. The company says its design also addresses a key limitation seen in some traditional MPC approaches: portions of signing that can occur outside protected hardware, which can complicate “fully offline” cold storage requirements.

Key takeaways

Blockstream’s new enterprise custody solution uses Thales HSMs alongside Secure-MPC to keep cryptographic signing operations inside hardware, per the company.
The architecture is designed to execute the full signing protocol within the HSM environment, unlike conventional MPC implementations that may rely on external processing.
Adam Back argues that complexity and heavy inter-device communication make MPC less suitable for truly offline custody in many designs.
Back cites the Binance wallet-signing incident as an example of how internet-connected systems and compromised credentials can expose signing infrastructure.
He also warns that AI-assisted adversaries may help attackers find subtle implementation errors in complex cryptographic systems.

Hardware-secured MPC built for enterprise custody

Blockstream’s announcement frames the offering as an institutional-grade custody workflow that combines two layers of security: secure hardware and cryptographic coordination. The company says Thales provides hardware security modules (HSMs), while Blockstream contributes its Secure-MPC technology.

The key engineering claim is about where the signing protocol runs. Blockstream says that, unlike certain MPC implementations in which parts of the signing process may happen outside the protected security environment, its architecture performs the entire signing protocol within the HSMs. That matters for institutions that want cryptographic operations to be insulated from broader system exposure—particularly when custody infrastructure is being operated at scale across many users, servers, and operational workflows.

By shifting the most sensitive computation into a dedicated security boundary, Blockstream aims to reduce reliance on perimeter controls and general-purpose systems, which can be undermined if attackers gain access to operational credentials or internal networks.

Why Adam Back prioritizes “simpler” offline security

Speaking to Cointelegraph, Blockstream co-founder and CEO Adam Back—known for his earlier Hashcash work that helped inspire Bitcoin’s proof-of-work—directly criticized how many MPC deployments have been implemented. In his view, MPC protocols are often too complex and too interactive to fit well with true offline cold storage.

Back’s core point is practical: “When people talk about MPC, they’re talking about some very esoteric and complicated protocols that are hard to secure and are very interactive. So you can’t really use them in a cold storage setting,” he said. The underlying argument is not simply about whether MPC is theoretically secure, but whether common real-world designs are workable when security models assume devices remain offline or minimally connected.

Blockstream’s new product can be seen as an attempt to bridge that gap: deliver the operational coordination benefits associated with MPC, while moving the signing computation into hardware that supports stronger isolation. For investors and custodians, the practical question is whether this approach reduces the attack surface that typically comes from complex protocol choreography across devices and systems.

Binance incident highlights risks beyond cryptography

Back also used a recent real-world breach scenario to underscore his perspective that custody failures frequently involve access and infrastructure compromise, not only mathematical weaknesses. He pointed to the Binance hack involving $6.3 billion as an example of what can go wrong when wallet signing infrastructure sits behind an internet-connected operational stack.

According to Binance’s investigation as described by Cointelegraph, attackers compromised employee devices and obtained credentials that enabled them to access the exchange’s wallet-signing infrastructure. Back used this as a cautionary example: even with firewalls and other controls, connecting systems to the internet creates pathways for unauthorized access, and attackers may leverage stolen logins to reach sensitive signing components.

If you connect equipment to the internet, even if it’s via firewalls and so on, there is risk that you can get unauthorized access. You can get hackers into the network, get hackers with access to logins they shouldn’t have and so on.

From an institutional standpoint, the takeaway is that custody security is not purely a cryptographic question. It also depends on operational design: what gets connected, which credentials can unlock access, and how much damage an attacker can cause once they penetrate the environment. Hardware-secured custody models aim to limit how far an intruder’s access can go once inside.

AI could raise the stakes for complex MPC implementations

Beyond today’s attack methods, Back argued that artificial intelligence could amplify existing security challenges. He suggested that AI tools may help adversaries identify small implementation errors that might otherwise go unnoticed in complex MPC systems.

“The AIs can help attackers find small implementation errors and things like that, which may not be obvious,” Back said. While this is an argument about future risk rather than a specific newly observed vulnerability, it aligns with a broader industry concern: as tooling becomes more capable, the cost of discovery and exploitation of subtle issues may fall—particularly in complicated cryptographic software where edge cases can be difficult to reason about.

In that context, Back’s position is that institutions should favor security architectures that reduce complexity where possible. He told Cointelegraph he hoped the broader custody industry would adopt offline-oriented security models, arguing that organizations should “adapt to the environment” by moving assets into cold storage to limit exposure to AI-assisted attack capabilities.

“Complexity is the enemy of security. And going back to basics is the answer, you know, just cold-store it,” he added.

What to watch next

Blockstream’s collaboration with Thales and the claim that the signing protocol executes entirely within HSM-protected environments will likely become a focal point for institutional buyers comparing custody designs. The next thing to watch is how widely this “hardware-contained signing with Secure-MPC” model gets adopted—and whether the industry converges on architectures that keep cryptographic operations isolated even as adversaries increasingly target the operational layers around custody systems.