VeilNet
News

The third generation of VeilNet: post-quantum from the root up

We started building on NIST's new post-quantum standards in February 2025. The third generation is here: a post-quantum PKI tree for identity, a gossip mesh keyed with ML-KEM, AI-driven routing and a high-performance Go core, open for evaluation now.

The third generation of VeilNet is out as a public pre-release, on seven platforms across five operating systems. One command puts a machine on a post-quantum network. There's no account to create, no keys to manage, and nothing of ours left in the loop.

Post-quantum from day one

In August 2024, NIST published its first post-quantum cryptography standards: FIPS 203 for ML-KEM and FIPS 204 for ML-DSA. In November, it set the clock, proposing that today's public-key algorithms be deprecated after 2030 and disallowed after 2035. Three months later, in February 2025, VeilNet started building a network on the new standards, making us one of the first teams anywhere to do it from the ground up rather than retrofit them onto something older.

Three generations later, identity, authentication and encryption are all post-quantum, from the identity of each machine to every packet it sends. Each generation removed another dependency on us. This one removes the last.

What's inside

A post-quantum PKI tree

Identity and authentication are a chain of ML-DSA signatures, from the organisation's own root key down to every machine. A device proves who it is before it sends a single byte, and every peer checks the chain for itself. The default suite is NIST category 5, the highest there is: ML-DSA-87 and ML-KEM-1024. How the tree works

A gossip mesh, keyed with ML-KEM

There's no central server. The machines in a network manage it themselves, spreading orders and knowledge to each other by gossip. Every link is keyed with ML-KEM, and every payload is encrypted a second time, end to end, with AES-256-GCM. Nothing that crosses the network can be harvested now and read later. How gossip works

AI-driven routing

Finding the best path through a mesh that changes by the second is a hard problem, and it gets harder with every machine you add. VeilNet uses AI to solve it. Routes are weighed continuously from the traffic already flowing, with no probes, and traffic moves off a link that is starting to degrade before it fails. How routing works

A high-performance Go core

The engine is written in Go and built for speed. On our test host, a direct round trip takes 0.79 ms, and a single stream runs at 92% of what one CPU core can encrypt with AES-256-GCM. Post-quantum security doesn't have to cost performance. See the numbers

Sovereign by design

Much of the security industry answered “stop trusting the network” by asking customers to trust a vendor instead: a console, an identity service, a cloud control plane. The third generation of VeilNet has none of them. There's nothing to sign into, no service of ours behind the scenes, and nothing at our end that could be breached, subpoenaed or switched off. The network belongs entirely to the organisation that runs it.

Try it today

The pre-release is open to everyone, and the quick start takes a few minutes. We've called it a pre-release on purpose: we'd rather find the rough edges with you now than after we call it done.

Planning a migration for a whole organisation? Request an evaluation, or write to sovereign@veilnet.net. More soon.