Overview
Scrydon empowers organisations with agentic AI, analytics, and data spaces designed around European data sovereignty requirements. The stack is rooted in open source and supports air-gapped, on-premise, and cloud environments.
The positioning targets regulated and public-sector organisations that need agentic AI capability without sending data through US-controlled platforms.
Leadership & org chart
In the news
- Not begging… networking.” 👏 Today I was at the European Commission to introduce Scrydon to Executive Vice-President Virkkunen’s cabinet: who we are, what we build, and how young European tech companies can compete in public procurement. A great and fruitful conversation. But the reason for this post happened outside the building. Two recent graduates were standing in front of the Berlaymont with these posters.👇 They asked if I could take a photo of them there, and then asked for career advice. One person’s advice only goes so
- No shots were fired, but still hundreds of French supermarkets lost their food and was there so much impact in Belgian supermarkets. Sunday night, someone hacked refrigeration controllers of Intermarché, Super U, E.Leclerc and Carrefour stores to defrost all the food, creating chaos, food waste and a lot of monetary impact. At the same time, Ahold Delhaize pulled systems offline across Belgium and the Netherlands. Is this what sabotage looks like now? The response to this needs to be national and fast! Today it's a supermarket, but
- 2 dead, 11 rescued." That was the first casualty figure many people read for the ALFA WATAN, sunk by drones in Bulgarias waters last night. It belonged to a different ship. Her crew is still missing. This morning we put 10+ open sources side by side and built one 3-page brief (attached). The sources disagreed on casualties, on distance offshore, and on who launched the drones. The brief shows every disagreement and which figure we used. At least seven governments are being briefed on these strikes. Put a wrong figure in a
- Drone strike, 70 nm off Bulgaria. At around 03:00 today, sea and aerial drones hit two cargo ships inside Bulgaria's exclusive economic zone. It is the first reported attack of its kind in those waters. The ALFA WATAN sank. Her crew is still unaccounted for. All 18 crew of the ABLE were rescued by the Bulgarian ferry DIOSCURIA. Two are in hospital in Varna. All of this, when sixteen hours earlier, a drone already sank the ROYAD MAMMADOV off Romania, killing two people. This means that we are currently at three merchant ships. Two
- A sandbox decides what an agent can reach, but it cannot decide what an agent is actually allowed to do—and treating them as the same thing is where most AI security models fail. NVIDIA just launched OpenShell (an Apache-2.0, Rust-based governed runtime for autonomous agents), and as builders of an enterprise agent runtime in Rust ourselves, we immediately tore down the codebase (v0.1.2) to see how it stacks up against production reality. We decided as Scrydon to put it to the test and check into: - Where the trusted authority sits:
- A sovereign AI company just joined two American startup programmes. Contradiction? No. Scrydon was accepted into NVIDIA 𝗜𝗻𝗰𝗲𝗽𝘁𝗶𝗼𝗻 this month! For a company that sells sovereign AI to defence, government and critical infrastructure, that deserves an explanation. Here it is. Dependency is not dependence. A dependency is something you buy from a supplier. Once it is yours, you run it, and nobody has to keep agreeing for it to keep working. Dependence is when it only works for as long as someone else keeps consenting. Take the
- A sovereign AI company just joined an American startup programme. Contradiction? No. Scrydon was accepted into NVIDIA 𝗜𝗻𝗰𝗲𝗽𝘁𝗶𝗼𝗻 this month! For a company that sells sovereign AI to defence, government and critical infrastructure, that deserves an explanation. Here it is. Dependency is not dependence. A dependency is something you buy from a supplier. Once it is yours, you run it, and nobody has to keep agreeing for it to keep working. Dependence is when it only works for as long as someone else keeps consenting. Take the
- AI safety went from AI terminators to much more real sandbox escapes. Looking forward in giving you an overview in what happened at the Frontier labs this Summer and what Scrydon is doing to counter.
Related profiles
Something wrong or missing? Send an update. Fixed within 24 hours.