Deeploy's EIC Accelerator Governs the AI a Bank Cannot See

Most AI governance work stops at documentation, which tells you what a system was supposed to do. Deeploy, founded in Utrecht in 2020 as a spin-off of the consultancy Enjins, sells governance that runs where the model runs: built-in explainability methods, drift and performance monitoring with alerts, bias and risk detection, automated compliance documentation for the EU AI Act, and an inventory that discovers every AI system and agent in an organisation, with the data staying inside the customer's own infrastructure.

The EIC Accelerator selected the project at its October 2024 cut-off for blended finance under the challenge for human-centric generative AI made in Europe, and CORDIS records a grant of €2,490,000 from March 2025 to February 2027 to scale the platform in the Netherlands and Germany and then across Europe. The company announced the award as up to €7.5 million including conditional equity.

Named customers, which is rare

Unlike most AI companies in this cut-off, Deeploy publishes who uses it. Its success stories name the Dutch bank bunq for AI compliance, the pension manager PGGM for a central AI overview, the cybersecurity company Hadrian, which it says now deploys AI models 80% faster, the mental healthcare provider Niceday, the clinical AI company Healthplus.ai for explainability, the textile manufacturer Vlisco for image recognition, and an insurer for a feedback loop. The regulated sectors it targets, banking, insurance, pensions, healthcare and government, are the ones where an unexplained model is a legal problem rather than a technical one.

Its own claims about scale are modest and specific: the platform serves millions of explanations a day for customers, and the product pages are backed by public documentation, tutorials, an open-source explainer library on npm and a GitLab organisation.

What the grant is paying for

The project's goal is to put real-time risk management, compliance and explainability inside the infrastructure where models run, aiming at GDPR and EU AI Act compliance through continuous monitoring and documentation, and reducing bias. The CORDIS page repeats the company's claim to be the "world's first human-centric AI MLOps platform", which nothing on the page verifies.

What has shipped since the grant started is a run of governance features rather than one big release: approval rules for deployments in November 2025, custom governance controls in October 2025, scheduled AI use case reviews in January 2026, NIST AI risk management framework and AIUC-1 mappings in February, a governance dashboard in March, an MCP server for AI agent governance in May, the EU AI Act Hub in July and a control framework whitepaper in August. CORDIS shows no periodic report.

1. Platform in production at customersYesNamed success stories including bunq and PGGM.
2. Agent governanceShippedMCP server launched May 2026.
3. Expansion to Germany and beyondNot reportedNamed in the project objective; no German customer announced.
4. EIC Fund equityNot announcedThe €7.5M package included conditional equity.

A seed round and an EU package

Deeploy raised a €2.5 million seed round in June 2023 led by SI3 Fund with Curiosity VC, Bonsai Partners, Emilia Capital and the angel investor Anton Loeffen. The EIC award, announced on 18 February 2025, is worth up to €7.5 million, of which €2.49 million is the grant recorded in CORDIS; the rest is conditional equity that has not been announced as invested. No further round is public and the company's accounts were not obtained for this article.

CEO Maarten Stolk framed the award in political terms: "Europe has set a global standard for AI ethics and governance, and we are proud to be at the forefront of this movement." The announcement argues that without proper controls, European organisations will depend on mostly American big tech.

Selling compliance as the product

Deeploy is one of the companies whose market the EU created. The AI Act, GDPR, the NIST risk framework and ISO 42001 all demand records of what a model does, and the company has built both a product and a content operation around them, including a public hub tracking the AI Act and a comparison of governance tools. The EIC's human-centric generative AI challenge is a close fit, and the grant is explicitly aimed at expanding the platform's explainability and at European AI sovereignty.