Update for the LaTeX bug : Inspiration
In less than 100 days, the EU AI Act becomes enforceable for high-risk AI systems across finance — from credit scoring to biometric KYC and insurance pricing.
We observed a critical disconnect: AI is built by technical teams, but approved by compliance and risk officers who lack visibility into how these systems behave. Meanwhile, insurers price risk without standardized, auditable signals.
Integreat was born from this fragmentation — to create a shared language of trust between builders, regulators, and insurers.
⸻
What it does
Integreat is a trust passport for AI agents, designed specifically for regulated financial institutions. It unifies three essential layers:
- Trust Score (0–100) — a composite, interpretable index:
- Technical robustness (45%)
- Legal compliance (35%)
- Ethical & societal posture (20%)
- AI Assurance Report — a standardized, auditable document, ready to be presented to regulators such as the ACPR
- AI Liability Coverage — distributed via Munich Re and Lloyd’s coverholders, priced using actuarial logic: Premium = Risk × Damage
Integreat closes the loop between evaluation, compliance, and insurance — in one platform.
⸻
How we built it
We designed Integreat as a modular platform combining:
- Evaluation pipelines for AI systems (robustness, bias, explainability)
- Regulatory mapping aligned with EU AI Act requirements
- A scoring engine translating technical metrics into business-readable signals
- Automated report generation for audit and compliance workflows
- GCP-based infrastructure for scalability and security
Our architecture bridges technical outputs and regulatory expectations — making AI systems interpretable across stakeholders.
⸻
Challenges we ran into
- Bridging disciplines: translating complex AI metrics into compliance-ready insights required aligning technical, legal, and insurance frameworks
- Standardization: no existing benchmark unifies evaluation, regulation, and insurability — we had to define our own abstraction layer
- Trust design: making the Trust Score both rigorous and understandable to non-technical stakeholders
⸻
Accomplishments that we’re proud of
- Designed a first unified Trust Score tailored to EU regulatory requirements
- Built an end-to-end loop from AI evaluation to insurance pricing
- Positioned Integreat in an empty market quadrant — no competitor combines compliance, evaluation, and insurance for EU fintech
- Defined a clear go-to-market: 500 mid-market fintechs, with 50 target design partners in Year 1
⸻
What we learned
- Trust in AI is not just a technical problem — it is organizational and financial
- Regulators, builders, and insurers need a shared interface, not separate tools
- Simplicity (a single score) is critical to drive adoption across stakeholders
⸻
What’s next for Integreat
- Onboard first design partners among EU fintechs
- Deepen integrations with regulatory bodies and audit workflows
- Expand insurance partnerships and refine actuarial models
- Extend the Trust Passport beyond fintech to other high-risk AI sectors
Log in or sign up for Devpost to join the conversation.