Govern installed in week one
Inventory, access policies and the activity log for every AI tool and connector, on your infrastructure or an EU tenant. €500 per month for six months.
Apply to the programmeEnterprise AI you control, and can insure.
AssurAI is the AI governance platform for Germany and the EU. Govern every model, agent and connector from one control plane, deploy private models inside your own perimeter, and turn the evidence into a certificate and AI-liability cover from a licensed partner carrier.
AssurAI GovernOne control plane for every AI system in the company
AssurAI AssureA certificate and cover on the same evidence
Built for the regulatory basis of Germany and the EU
EU AI ActEN 18286:2026ProdHaftG 2026GDPRISO/IEC 42001NIST AI RMFBNetzA · KI-MIG
AI for the whole company
Govern AI without slowing anyone down.
Monitoring agents built on our own trained model watch every thread of AI activity, surface misuse, bad actors and threats, and bring your team only the handful of decisions that need a person. The same people, the same hours, with AI inside the perimeter.
Explore how it works
Purpose-built for governed AI
Three products. One control plane.
AssurAI Govern
Every AI system in the company, visible and under control.
A lightweight gateway on each workstation and server routes AI and connector traffic through one point you control. Our own trained model inspects every call for prompt injection, data exfiltration and unsafe tool use before it executes.
- Living inventory of models, agents and MCP connectors, including shadow AI
- Access policies enforced at the gateway, per tool, system and data class
- Approved skills and configurations deployed centrally to every team
- Usage, cost, incidents and evidence packs mapped to EN 18286 and the AI Act
Available now · self-hosted or EU-hosted
On-premise or dedicated EU tenant · zero data egress
AssurAI Private
Your own private models, deployed inside your perimeter.
For organisations that cannot send data to a public AI service at all. Open-weight models run in your data centre or a dedicated EU tenant, fitted to your use cases, shipped with the gateway and threat protection already attached, and operated for you.
- No prompt, document or log leaves your infrastructure
- Models selected and tuned for document review, customer service, engineering
- Governed and insurable from the first call
- Managed operation: updates, evaluation, capacity, incidents
AssurAI Assure
A certificate and AI-liability cover on the same evidence.
Continuous evidence from Govern becomes an AI risk score and a one-page Certificate of AI Assurance mapped to EN 18286. A licensed partner carrier underwrites named-peril cover on that evidence, inside one annual agreement. AssurAI never holds risk.
- Named perils: wrong outputs, agent action errors, data leakage, bias, IP, regulatory defence
- Third-party liability, first-party operational loss, defence costs
- Better score, lower premium at renewal
- The certificate answers the enterprise AI risk review on page one
With partner carrier · cover subject to carrier wording
The AssurAI model
Our own trained model for governance and threat protection.
Rules catch what you already know. The AssurAI model, trained in-house on governance and attack patterns, runs at the gateway and decides in milliseconds whether a call is safe, risky or an attack. Tuned on your estate; nothing leaves the EU, nothing is shared between customers. Defensive only.
Explore the model
Monitoring agents
Observe. Evaluate. Surface.
Agents follow every thread of activity through the gateway. The model classifies each event against your policies and known attack patterns. Routine traffic is logged, threats are blocked, and the few cases that need judgement reach the right owner with the evidence attached. On a typical day: a few held actions, one or two incidents, a weekly summary.
See how the agents work
Solutions
Governed AI across industries.
Financial services
Decision systems and customer AI under BaFin, the AI Act and the new product-liability regime.
- Credit and claims models
- Advisory chatbots
- Vendor AI oversight
Manufacturing
AI in engineering, quality and the supply chain, with trade secrets that cannot leave the plant.
- Private engineering assistants
- Quality inspection models
- Supplier document review
Professional services
Client data in every prompt, and client contracts that now ask how AI is governed.
- Document and contract review
- Client-facing assistants
- Certificate for RFPs
Software and AI vendors
Selling AI to enterprises that run a security review before every contract.
- Certificate for security reviews
- Insured outputs as a sales argument
- Evidence for EN 18286
Design partner programme
Why early teams work with AssurAI.
Shape the certificate and the cover
A fortnightly product review, an AI risk snapshot for your board and customers, and first access to cover when the carrier goes live.
Talk to the founding teamFrom 2027: one annual agreement
Govern, Private and Assure for all AI systems in scope, €20,000–50,000 per year priced per system and cover limit. For comparison, ISO 42001 certification alone costs €130,000–250,000 in year one.
See pricing
The latest
The regulatory basis changed in autumn 2026.
14 October 2026
First harmonised standard under the AI Act
EN 18286:2026 was published in the Official Journal. Conformity carries a presumption of compliance with the AI Act's quality-management and documentation obligations.
21 October 2026
Carriers confirm the gap
Chubb's chief executive told investors that AI liability is "not contemplated in casualty underwriting" and that certification standards are needed to underwrite it. No German carrier offers an AI-liability product.
9 December 2026
Strict liability for software and AI in Germany
The revised Product Liability Act takes effect. Courts may order disclosure of technical documentation and presume a defect when it is withheld.
Security and sovereignty
Your data stays in your jurisdiction.
Self-hosted or EU-hosted. Prompts, logs and evidence never leave the EU. The control plane is built on an MIT-licensed open-source core that your security team can audit line by line; our work is the evidence model, the risk score, the trained model, the certificate and the carrier integration.
Learn moreEU AI ActReg. 2024/1689
EN 18286harmonised 2026
GDPREU data residency
ISO/IEC 42001mapped controls
ProdHaftG 2026§§19–20 evidence
Open coreMIT licence
Questions
Frequently asked.
What is AssurAI?
AssurAI is an AI governance platform for organisations in Germany and the EU. It inventories and controls every AI model, agent and connector, deploys private models inside your perimeter, and turns the evidence into a Certificate of AI Assurance and AI-liability cover.
Is AssurAI an insurance company?
No. AssurAI does not hold risk. AI-liability cover is underwritten by a licensed partner carrier on the evidence AssurAI produces, and is subject to the carrier's policy wording.
Which regulations does AssurAI help with?
The EU AI Act and its 2026 Omnibus, the harmonised standard EN 18286:2026, the German Product Liability Act 2026, DORA, GDPR and ISO/IEC 42001.
Where is our data stored?
AssurAI runs self-hosted or in a dedicated EU tenant. Prompts, logs and evidence stay in the EU and are never used to train for other customers.
Does AI governance slow teams down?
No. Approved use passes straight through the gateway. Monitoring agents built on the AssurAI model handle routine traffic and surface only the few decisions that need a person.
How much does AssurAI cost?
The design partner programme costs EUR 500 per month for six months. From 2027 an annual agreement covering Govern, Private and Assure costs EUR 20,000 to 50,000, depending on scope and cover limit.
Ready to put AI to work, safely?
See Govern running on sample data and walk through the certificate and cover structure with the founder. Reply within two working days.
Request a demo
Company
Talk to us.
Organisations, carriers, brokers and investors reach the founder directly.
hello@assurai.example
Berlin, Germany