The spin-off Validaitor develops methods for reliably testing AI applications, thereby building trust in a technology that increasingly shapes our lives.
Artificial Intelligence (AI) promises efficiency and progress yet many people feel uneasy about working with it. We get into an autonomous taxi and trust that the vehicle knows the way and takes us safely to our destination. Companies use AI powered tools to review job applications and hope that the algorithms will find the right person while acting fairly and objectively. In both cases, however, it remains unclear how exactly these decisions are made. Studies show that mistrust is especially high in sensitive areas such as human resources, autonomous driving or medical applications. Who guarantees that a system won‘t act differently under pressure? The KIT spin-off Validaitor tackles this very mistrust. Rather than relying on blind faith, the founders‘ software works like airport security checks: it scans the AI, checks algorithms for robustness, fairness and reliability, among other things, and thus makes the „black box“ more transparent.
The testing dashboard visualizes test runs, benchmarks and quality metrics for an AI model — for example, users can select and run tests or view results immediately.
Operationalization instead of certification
„When I first heard about the risks associated with Artificial Intelligence, I immediately realized that this was a very serious matter. My first thought was: Maybe I should develop a toolkit to make it easier for development teams to test AI,“ recalls founder Yunus Bulut. Within six years, a personal concern turned into a toolkit, the toolkit turned into a platform and the platform turned into a KIT spinoff: Validaitor. What drives him and his team is not pure tech fascination, but the question of how trust can be built in a technology that often works in the dark. When algorithms decide about people in areas such as medicine, mobility or human resources, trust alone is not enough – controllable, reproducible checks are needed. Numerous proposals, guidelines and even laws already exist, but they often remain abstract. Validaitor aims to close this gap by offering verifi cation processes not as one-off audits, but as an integral part of everyday development and operations. „It‘s not just about possible external certifi cation, but about operationalizing these quality assurance processes directly within the company,“ says Cofounder and CRO Gert Paczyński and Chief Revenue Offi cer. „We bring developers, compliance and risk teams together in one platform, standardize workfl ows and integrate tests directly into the development process.“
How Validaitor works
Artificial intelligence makes more and more decisions in our everyday lives, but often remains a black box for outsiders. Without transparent, traceable testing, the risk of wrong decisions, misuse and loss of trust rises – especially for companies that lack inhouse testing expertise. Validaitor provides an integrated platform that automatically assesses the risks of AI projects, performs appropriate tests and documents the results transparently in accordance with applicable regulations (e.g. the EU AI Act).
The Platform as a Control Hub
Validaitor‘s solution is an integrated platform that combines test tools, governance modules, risk-management features and a public trust center on a single interface. Users can select and run commands within a single platform without having to stitch together multiple disparate tools. The workflow is deliberately simple: Users register a project and provide metadata, such as the intended use, deployment context or affected stakeholders. The platform automatically suggests appropriate, risk-aligned tests and supports their execution – covering documentation, workflow integration and linking test results to compliance requirements. It includes dedicated Compliance Assessments that take into account regulations such as the EU AI Act and AI‑specific ISO standards. In addition, it offers management functionalities such as an AI model/system registry, enabling traceable versioning and metadata tracking of machine‑learning models. Beyond the technology, the platform saves time: automation eases the ever‑tightening development resources, streamlines testing and prevents companies from losing valuable time in the fast‑moving AI landscape.
The Validaitor Team: Cem Daloglu, Michael Graf, Yunus Bulut, Gert Paczynski und Sebastian Krauss (f.l.t.r.)
From Proof of Concept to Standard
What started as a project has now reached market readiness: With the help of funding from the Federal Ministry of Research, Technology and Space (BMFTR), support from the KIT-Gründerschmiede and an initial venture-capital round, the team has brought the platform to market maturity. Validaitor has been operating as a company since September 2024, has expanded its headcount to a team of 11 and already counts its first paying customers – from SMEs and government agencies to forensic authorities in Germany. In the second funding phase, „Startup Secure“, the spin-off plans to scale further, refine its product and expand market access, aiming to be fully compliant with the upcoming regulatory requirements slated for mid‑2026. The strength of Validaitor lies in its promise: trust is important, control makes it resilient. For many small and medium-sized enterprises that lack dedicated AI security teams, this means more specific safety – compliance becomes manageable, tests repeatable and results transparent. In the short term this protects against poor decisions and legal risks. In the long term the integrated approach could become a widely adopted practice that makes AI decisions comprehensible to users and the public – not just as an academic ideal, but as an everyday reality. And that, according to Yunus and Gert, is exactly what society needs right now.
Diese Seite nutzt Website-Tracking-Technologien von Dritten, um ihre Dienste anzubieten. Ich bin damit einverstanden und kann meine Einwilligung jederzeit mit Wirkung für die Zukunft widerrufen oder ändern.
