We are pleased to announce our investment in Wodan AI, a Spanish-Belgian deep-tech company building the infrastructure layer for secure and private AI operations. 

The €2 million pre-seed funding round was led by JME Ventures, Swanlaab, and Adara Ventures, with participation from ScaleFund.

The challenge: AI adoption is outpacing AI security

AI adoption across enterprises is accelerating rapidly, particularly in highly regulated and data-sensitive sectors such as financial services, healthcare, defence, and public administration. 

However, most AI systems today still require data to be decrypted during training or inference, exposing sensitive information to infrastructure providers, model operators, and attackers.

This creates a growing and systemic cyber risk. Nearly all organisations rely on third-party providers, and AI inference platforms handling sensitive data have increasingly become high-value attack surfaces. 

At the same time, new regulations, such as the EU’s Digital Operational Resilience Act (DORA), explicitly mandate stronger protections, including the encryption of data-in-use.

Introducing Wodan: AI that runs on encrypted data

Wodan is addressing this problem at its root. The company has developed a platform that enables Machine Learning, Computer Vision, and Large Language Models (LLMs) to operate directly on encrypted data, without ever decrypting it at any stage of the workflow.

At the core of Wodan’s technology is Fully Homomorphic Encryption (FHE). This allows data owners to encrypt their data locally, send it to an AI model for inference or training, and decrypt only the final result – ensuring that neither the AI provider nor the infrastructure ever sees the raw data.

Wodan’s platform supports:

• Converting existing models into encrypted-data-compatible versions
  
• Training new models directly on encrypted datasets
  
• Deploying secure AI workloads on any cloud or on-prem infrastructure
  
• Managing and monitoring encrypted models through a unified interface
  

The result is end-to-end protection for both data and algorithms, enabling AI adoption in environments where it was previously too risky to deploy.

An experienced founding team at the intersection of AI and enterprise security

Wodan was founded by Bob Dubois (CEO) and Manuel Pérez Yllán (CTO). Bob has spent two decades in enterprise sales and digital transformation roles at companies such as Red Hat, Commvault, and Filecoin, while Manuel is a serial AI entrepreneur, machine learning expert, and professor with deep experience in cryptography and applied AI systems. 

This combination of commercial execution and deep technical expertise positions Wodan well to tackle one of AI’s most complex infrastructure challenges.

A platform vision for secure AI at scale

Wodan’s ambition extends beyond point solutions. The team is building what they describe as the “Hugging Face of secure models” – a platform where organisations can train, deploy, and share AI models securely, while preserving data privacy and regulatory compliance.

The company is initially focused on financial services, where AI adoption is high but security constraints are particularly acute. Early traction includes pilot projects with financial institutions, with plans to expand into healthcare, legal, defence, government, and other sectors that handle highly sensitive data. 

This investment will allow Wodan to strengthen its team and technological infrastructure, as well as position the company as a leading European player in secure and private AI, an increasingly strategic area for the European Union’s digital autonomy.

The Adara view: Fully Homomorphic Encryption as critical AI infrastructure

At Adara, we believe Fully Homomorphic Encryption is no longer a theoretical breakthrough. It is becoming a commercially relevant technology for addressing one of the most urgent risks in modern AI systems.

By enabling ML, LLM, and Computer Vision models to operate on encrypted data instead of decrypted data, Wodan makes it possible to deploy AI with end-to-end cybersecure operations. This fundamentally changes the threat model for AI. Even if infrastructure is compromised, sensitive data and model interactions remain protected.

Recent cyberattacks on inference providers and AI platforms handling sensitive data underscore how exposed today’s AI stack has become, and why this trend will only accelerate as AI adoption grows. We view Wodan as an infrastructure-level solution to this expanding attack surface, addressing security at the point where it matters most: data-in-use.

Wodan’s approach aligns strongly with our thesis around deep, defensible technologies that enable safe deployment of AI in mission-critical environments. With strong technical foundations, regulatory tailwinds, and a clear platform vision, we believe Wodan is well-positioned to become a cornerstone of secure AI infrastructure in Europe and beyond.

We are excited to partner with Bob, Manuel, and the entire Wodan team as they build the next generation of cybersecure AI.

Learn more about Wodan: www.wodan.ai