This agent may order known items from approved suppliers.
COUNTERPARTIES
42 approved suppliers
LIST
ACTIONS
create_order · cancel_order
2 ALLOWED
MONTHLY BUDGET
€90,000
HARD
REVOCATION
by principal
IMMEDIATE
Principal · Rheinwerkv3 · active
MANDATE
✓€12,400 is within the €15,000 limit
✓Supplier is on the list
✓create_order is an allowed action
DECISION
ALLOWED
Executed at the endpoint: order placed.
RECEIPT · EVERY CALL, ALLOWED OR NOT
Model
which model acted
Agent
which agent
Organisation
what it belongs to
Responsible person
who owns it
Time
when it happened
Tools called
create_order, € 12,400
Mandate
which mandate applied
Decision and result
allowed, order placed
Call 1 was blocked and recorded too.
To quantify your risk level we collect loss data and continuously test your agents – so underwriters can price the premium.
INPUT
OUTPUT
Customer questionnaire
Industry, revenue, what the agent may do, how many actions, at what amount a human steps in
Sandbox / Arena
Error rate, loss per 1,000 actions, does the agent bypass safeguards
Mandate
Limit per action, limit per month
Calculator
Risk tier
for the customer
Data sheet
for the underwriter
Today’s safety systems are slow and probabilistic or need heavy workarounds.
01
OAuth
Broad. Not for cross-system workflows. Needs heavy workarounds.
02
LLM guards
Expensive, slow and probabilistic.
03
Human in the loop
Time-consuming, with a risk of fatigue.
04
Bypass paths
Direct tool access can bypass checks. With , agents hold no keys.
excels at safety, speed, and accountability.
Attack success
Percent · scale 0–100 · lower is better
No protection71.5%
OAuth15.1%
LlamaFirewall63.4%
Invariant24.0%
Progent on52.7%
Progent off5.4%
wiya C21.9%
Check time per case
Seconds · scale 0–16 · lower is better
No protection0 s
OAuth< 0.001 s
LlamaFirewall0.72 s
Invariant0.073 s
Progent on15.4 s
Progent off2.22 s
wiya C20.017 s
Total cost per case
US dollars · scale $0–$0.02 · lower is better
No protection$0.0078
OAuth$0.0078
LlamaFirewall$0.0079
Invariant$0.0084
Progent on$0.0186
Progent off$0.0093
wiya C2$0.0081
Banking damage
Euros per 1,000 calls · Banking only · lower is better
No protection€83,981
OAuth€37,008
LlamaFirewall€96,190
Invariant€3,663
Progent on€35,114
Progent off€13,389
wiya C2€26
AgentDojo (ETH Zurich) · Qwen3-235B. Four-world average (Travel, Slack, Banking, Office), except Banking damage. wiya C2 is the arena version. Human reviewer simulated; human waiting time excluded from check time.
The field is split into layers. Only joins them as a neutral player, with forensic evidence.
IDENTITY
Entra Agent ID
PRIMARY FOCUS
Identity, access and agent lifecycle
PLATFORMS & GATEWAYS
Zenity
PRIMARY FOCUS
Agent execution and runtime controls
AGENT-AUTHORITY TOOLS
Agent Authority ChainFidacyPraesidiumSecureAuth
PRIMARY FOCUS
Delegation, permissions and authority chains
RISK RATINGS & ASSURANCE
BitsightAIUCMount
PRIMARY FOCUS
Risk assessment, assurance and insurance
Selected examples from public documentation, October 2026. Categories overlap.
Why Europe? EU regulation makes accountability a must, not an option.
DORA, since January 2025: firms answer for every IT provider they use. AI Act, from December 2027: logging and human oversight.
20 %
of EU companies use AI, up from 13.5 % in 2024. Large companies: 55 %.
€44M
Germany, first customers: about 1,450 supervised firms at €30k a year.
€210M
EU/EEA regulated firms: about 7,000 supervised firms at €30k a year.
€1.6bn
Ceiling: about 53,000 large EU firms at €30k, if all run agents.
We start where agents already work but wait for a human, then move on to insurers and insurable proof.
NOW
Runtime control
WHO BUYS
Regulated firms with live agents.
THEY GET
A signed mandate, enforced on every call. A hidden email asking for €9,000 is blocked.
BOTTLENECK
A person approves every action by hand.
NEXT
Underwriting tools
WHO BUYS
Cyber underwriters at insurers.
THEY GET
A risk calculator: test results in, expected loss and premium range out.
BOTTLENECK
Insurers have no data on agent risk.
LATER
Insurable attestation
WHO BUYS
Firms that run agents, for their insurer.
THEY GET
Signed evidence from receipts: cover again, or a lower premium.
BOTTLENECK
Insurers do not yet accept this evidence.
PARTNER
EARLY SIGNALS
Conversations so far
PoC in work
SPARKBMDS
Permission audit in work
Non-disclosed digital broker
Risk-control interview
Mosaic
Underwriting interviews
msg opens the market: first the pilot, then regulated firms.
Design and implementation partner
msg gives us access to its network of banks, insurers and the public sector. msg implements, we license.
01 · BUILD WITH MSG
Internal pilot in November
Internal workflows built with msg’s security know-how. Goal: a working pilot with measured findings.
02 · SELL WITH MSG
Introductions to regulated firms
msg introduces banks, insurers, the public sector and more. Price test: €20–40k per firm per year.
Built at the intersection of AI, business, security and human systems.
Luc Robuste
Founder & CEO
Organisational psychologist, Law & Economics background, self-taught AI engineer. Strategy, product, fundraising.
Karol Genczyk
Co-Founder & CSO
Forward deployed engineer. Prototyping and custom software for customers. Builds integrations and adapts wiya to their workflows.
Aqib Amir
Co-Founder & CTO
M.Sc. CS, University of Bonn. Builds the SDK and adapters for the gateway. GenAI at Cogent Labs, backend at Grandeur. Data visualisation and simulation at DHL.
Burak Erdogan
Head of Security
Builds the gateway. Security engineer in regulated and security-critical environments, including NATO projects.
Han Jiadong
Machine Learning
Mathematician and ML engineer. Shenzhen, Paris-Saclay, Bonn. Builds a knowledge graph of loss data and trains machine learning models.
Fahad Memon
Risk Quantification
Economics at Bonn. Derivatives valuation at E.ON. Quantifies risk levels for companies and their AI agents.
ADVISORS
Sven Krueger
Enterprise AI, ex CMO T-Systems and adesso
Himanshu Rathi
Cyber & Tech E&O underwriter at Liberty
Gereon Neuhaus
Division Manager Industry, msg security advisors
Kersten Müller
Principal Business Consultant, msg
The ask: €450k for 12 months of runway, to prove three things.
RUNWAY
12 months planned
About €33k a month. The €50k reserve extends it to roughly 13 months. No revenue assumed.
PROOF 1 · PILOT
A working pilot
Mandates and receipts in one internal workflow with msg, with measured findings.
PROOF 2 · MARKET
A first paid licence
A regulated firm pays, and its risk control accepts our receipts as evidence.
PROOF 3 · INSURERS
A documented assessment
An underwriter writes down what our evidence is worth for pricing.
€210k
Team
€48k
Go-to-market
€57k
Compute & hardware
€85k
Operations & setup
€50k
Reserve
It all starts with four questions.
W<<WHO<WHOSE<WHAT<WHERE<<
01
WHO
Who is the agent?
A verifiable identity for the agent and the organisation behind it.
02
WHOSE
Whose behalf does it act on?
The legal entity and the responsible person who answer for it.
03
WHAT
What may it do?
The signed mandate: actions, limits and counterparties.
04
WHERE
Where does it act?
In which systems and tools. Every call is checked at the gateway and recorded.