EU AI Act
- Art. 50 transparency banner, locale-aware and customizable
- Art. 4 AI-literacy module, 6 modules, 22 questions, certificate
- Art. 26 deployer-obligations mapping
- FRIA + DPIA templates, pre-filled with a risk catalogue
SOC 2 Type IIGDPR CompliantLocal Deployment
Kaldryn One: The Local AI Management Platform Your Organization Needs.
Private AI that passes the audit. Kaldryn runs a ChatGPT-class assistant entirely on hardware you own, built for the control, security and compliance regulated industries demand, with no per-token fees, and not a single byte sent to the cloud.
Kaldryn
Private assistant · on your server
Summarise our Q3 sales report and flag the biggest risks.
Message Kaldryn…
How it works
One platform, running local AI for your whole company.
Kaldryn is an enterprise platform you deploy, not another app on every laptop. One private AI server inside your network, the whole team connects to it, and IT decides which models run, who uses them, and how.
- 01
You deploy the platform
Kaldryn installs on the GPU servers you already own, or arrives pre-configured on Kaldryn One, our appliance for companies without a data center or GPUs of their own. Either way, it runs inside your network, under your control.
- 02
Your whole team connects
Everyone works against the same private AI server, no local installs, no models scattered across laptops. Your documents are stored and indexed centrally, inside your network, and every answer is grounded in them.
- 03
You scale it, safely
IT approves the models, controls who can access what, and sees usage in one audit log, no shadow AI, no guesswork. Start with one team, grow to the whole company on the same platform.
Kaldryn One · online
A normal office. One quiet Kaldryn One, serving AI to the whole company.
No data center? That's what Kaldryn One is for.
One box
plugs in, no server room
Whole company
everyone connects to it
100% private
AI inside your own network
Kaldryn vs. Microsoft Copilot
Same capability. Opposite ownership.
Microsoft Copilot
Kaldryn
Cost model
Microsoft Copilot$30 per user, per month, forever
KaldrynOne-time hardware + setup, unlimited users
Monthly cost (30 users)
Microsoft Copilot~$900 / mo, rising with headcount
KaldrynElectricity
Token costs
Microsoft CopilotCopilot credits at $0.01 each, credits per prompt vary
KaldrynZero, no credits, no metering
Where your data is processed
Microsoft CopilotMicrosoft's cloud
KaldrynYour hardware, your building
Internet required
Microsoft CopilotAlways
KaldrynInstall only
Model choice
Microsoft CopilotMicrosoft decides
Kaldryn200+ open models, you choose
Fine-tune on your data
Microsoft CopilotNot offered
KaldrynBuilt-in Training Studio
Air-gapped operation
Microsoft CopilotNot possible
KaldrynDesigned for it
Pays for itself. Thirty Copilot seats run about $10,800 a year, every year. Most organisations recover the Kaldryn hardware investment within 3-6 months, then the meter stops.
Compliance, in the box
Self-hosted AI your in-house counsel can actually sign off on.
GDPR
- Art. 17 erasure that propagates across conversations, documents, embeddings, LoRA and backups
- EDPB 28/2024 LLM guidance integrated
- DPA templates with healthcare + DORA addenda
- Data residency = your hardware
Security posture
- Immutable audit log with SIEM export (syslog / HMAC webhook)
- Image-integrity monitor with hourly baseline and drift alerts
- One-click sanitized Security Evidence Bundle for CISO review
- NIST 800-53 Rev 5 readiness + SOC 2 CC6 / CC7 mapping
Sector overlays
- Healthcare, MDR / IVDR, Art. 9, EHDS
- Public sector, Art. 27 FRIA, Art. 49 registry, NIS2
- Finance, DORA, CRA, EBA, MiCA
- Sweden & Nordics, IMY, Dataskyddslagen, Offentlighetsprincipen
Not a legal guarantee. Kaldryn ships the tooling and the documentation scaffolding; your EU-qualified lawyer does the final review.
Deployment
Live in under 10 minutes. Prefer to run Kaldryn on servers you already own? No Kubernetes. No DevOps team. Two files and a server.
01
Download
A single signed, checksummed .run installer with every image, model and dependency baked in.
02
Run
Execute as root on a fresh Ubuntu box. 30 preflight checks, services stood up, the default model warmed. Zero manual config.
03
Activate
Paste your signed license. It verifies offline against the hardware fingerprint, and you share the chat link.
Bare metal, Proxmox, VMware or a cloud VPS, if it runs Ubuntu 22.04 or 24.04, Kaldryn runs on it. Pre-built GPU appliances available.
Who it’s for
Built for teams that can’t send their data to someone else’s cloud.
Growing businesses
Enterprise AI without enterprise costs. One flat appliance cost, unlimited users, and headroom past 500 seats with no license change.
- No per-user subscription fees
- Flat cost, unlimited users
- Scales without re-architecting
Security-sensitive industries
Healthcare, legal, finance and defence, where data cannot leave the building. Zero cloud connectivity after install.
- DLP, SSO/SAML, immutable audit
- SIEM export, real-time
- EU AI Act + GDPR kits included
Appliance customers
A pre-built GPU server with Kaldryn installed and a license bound to the hardware. Built for air-gapped sites.
- Pre-built GPU appliance
- OTA USB updates
- Time-boxed remote support tunnel
What we do
An AI specialist consultancy & product company.
Kaldryn is a specialist consultancy headquartered in Stockholm. We don't just teach organizations how to apply the principles of modern work, business applications and AI, we build and implement the solutions alongside them.
Our mission
Make modern technology simple, safe and genuinely useful for the people who work with it every day.
Experience across
Banking, public sector, healthcare, energy, real estate and manufacturing, across the EU and the Gulf.
9
Specialists
Senior technical consultants, including the founder
3
Practices
Modern Work, Business Applications, Artificial Intelligence
2
Platforms
Solutions in the Microsoft and Anthropic ecosystems
1
Goal
Qualitative and measurable outcomes in every initiative
From the Kaldryn library
Our latest research
Field notes, white papers and plain-language briefings on running AI on your own hardware, keeping data in the building, and staying on the right side of European rules.
White paper4 Jul 2026
Local AI in Sweden and the EU: The 2026 Guide to Sovereign Deployment
Why Swedish and European organisations are moving AI on-premise in 2026, the regulation driving it, who is deploying local AI, and how to choose a sovereign platform.
ReadEU regulation2 Jul 2026
The AI Act deadlines just moved: what the 2026 Digital Omnibus changed
In mid-2026 the EU pushed back the AI Act's high-risk deadlines. Here is what moved, what did not, and what deployers should actually do about it.
Read
Blog2 Jul 2026
Why we put a datacenter in a locker
The thinking behind Kaldryn One, why the future of business AI for most companies is a quiet black box in the corner of the office, not a bigger cloud bill.
ReadEU regulation1 Jul 2026
The EU AI Act timeline: what applies now and what's coming
A plain-language walkthrough of the AI Act's phased application dates, what organisations deploying AI must already comply with, and what lands next.
Read
Blog27 Jun 2026
Why European companies stopped trusting the cloud with everything
This did not come from nowhere. A court ruling, a foreign law, five years of unease and, in 2026, governments actually moving. A short history of the shift.
ReadEU regulation24 Jun 2026
The Cyber Resilience Act: security duties for anything with software
The Cyber Resilience Act sets baseline security rules for products with digital elements. What it requires, when it bites, and how it touches AI systems.
Read
Blog24 Jun 2026
The questions every IT lead asks before going on-prem with AI
Hardware, models, maintenance, updates and what happens when the internet goes down, the honest answers to the questions we hear in every Kaldryn evaluation.
ReadWhite paper24 Jun 2026
Sovereign AI: What It Means and Why Regulated Enterprises Need It
A plain-language primer on sovereign AI, what it is, how it differs from cloud AI, and why regulated organisations are bringing their models in-house.
ReadWhite paper20 Jun 2026
On-Premise AI: A Reference Architecture for Air-Gapped LLM Deployment
A practical reference architecture for deploying large language models on-premise and air-gapped, network topology, GPU sizing, model serving and zero-egress design.
ReadWhite paper18 Jun 2026
Digital sovereignty in the EU: a 2026 field guide
Europe spent 2025 and 2026 waking up to its dependence on a few foreign cloud providers. A field guide to the laws, the money and the moves that follow.
ReadEU regulation18 Jun 2026
GDPR and self-hosted LLMs: the questions your DPO will ask
Lawful basis, minimisation, the right to erasure and embeddings, DPIAs, how the GDPR conversation changes when the model runs on your own hardware.
ReadWhite paper16 Jun 2026
The EU AI Act and Self-Hosted LLMs: A Compliance Checklist
How the EU AI Act maps onto self-hosted and on-premise LLM deployments, a practical, plain-language checklist for legal, compliance and security teams.
Read