FIS research program: On the trail of AI fraud
With the signing of the contract on December 9, 2025, the Agentur für Innovation in der Cybersicherheit GmbH (Cyberagentur) is launching the multi-year research program “Forensics of Intelligent Systems (FIS)”. Three selected contractors will start their work to develop new methods so that intelligent systems can be analyzed and checked for manipulation in a way that can be used in court in the future.
An AI-supported application system makes automated pre-selections by prioritizing CVs according to certain characteristics. But what happens if manipulation of the training data ensures that applicants with certain backgrounds are systematically disadvantaged? Or if the system suddenly makes significantly different decisions after an update? The black box logic of many AI applications makes such effects barely comprehensible – with potentially far-reaching consequences for companies, society and the rule of law.
This is where FIS comes in. Digital forensics provides the methodology for securing and analyzing digital traces of intelligently networked systems and evaluating them in court. However, there is still a lack of standards, tools and procedures to detect and prove manipulation of changing, learning AI systems beyond doubt. The “Forensics of Intelligent Systems (FIS)” research program aims to close this gap and create a completely new quality of digital intelligence capability.
With FIS, the Agentur für Innovation in der Cybersicherheit GmbH (Cyberagentur) is bringing together three teams made up of renowned research institutions, industry partners and legal expertise. The aim is to develop procedures and tools that make the behavior of AI systems transparent, reconstructable and reliably verifiable.
“Today, intelligent systems determine security, mobility and social processes. With FIS, we are creating a research basis for the first time that makes it possible to make AI not only powerful, but also forensically verifiable, legally secure and therefore trustworthy,” says Prof. Dr. Christian Hummert, Research Director of the Cyberagentur.
It was therefore important to the Cyberagentur researchers to commission different approaches to solving the research program and to fund their work:
INSIDE – Intelligent text and audio processing
The research project “INtelligent Security Inspection of Data using forensic Extraction and Analysis (INSIDE)” aims to create the basis for the forensic auditability (“forensic readiness”) of future AI systems. The focus is on the development and testing of innovative forensic methods to be able to systematically, comprehensibly and legally prove manipulations, attacks and anomalies in text-based AI applications – especially in the area of cyber security and in the analysis of network and text data.
INSIDE addresses both technical and legal challenges: Forensic procedures are developed that not only reflect the current state of the art, but also meet the requirements for evidence and compliance in accordance with applicable EU regulations (e.g. AI Act).
The aim is to strengthen the traceability, integrity and transparency of AI systems and thus create a resilient basis for technical sovereignty, legal certainty and trustworthy artificial intelligence in the digital space.
The project brings together leading experts in artificial intelligence, cyber security, digital forensics and IT law from the Technical University of Mannheim, the Institute for Security and Safety, the University of the Federal Armed Forces Munich, the University of Twente, cyberintelligence.institute GmbH (CII) and the FZI Research Center for Information Technology.
EMAS-KI – Intelligent image and video processing
The consortium around Atos, the Fraunhofer Institute for Applied and Integrated Security AISEC, the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, the Institute for Internet Security at the Westphalian University of Applied Sciences and the University of Cologne is working on the EMAS-KI project to develop court-proof methods for continuously learning image and video AI systems.
The research project, for which the parties involved have qualified since 2023 with a preliminary short concept and long concept, is developing prototype methods and tools that should enable the court-proof detection of manipulation of continuously learning AI systems. To this end, a simulation environment is being set up in a research laboratory in which software for forensic methods can be tested. Existing AI models will be exposed to simulated attacks in order to investigate which legally usable traces can be secured. In addition, AI algorithms will be adapted to enable early detection.
FORAImobility – Forensics of autonomous systems
ZF Friedrichshafen AG is leading the FORAImobility consortium together with the “Law and Digitalization” research initiative of the Center for Human | Data | Society at the University of Konstanz. The aim is to develop innovative methods to detect manipulations and attacks on AI-based autonomous systems in the field of mobility and to be able to prove them in court.
Autonomous systems such as self-driving vehicles are highly complex and will continue to learn in the future. To this end, they must be securely protected against new types of cyber attacks. The project addresses these challenges by developing concepts for “forensic readiness” and “forensic-by-design”, i.e. the ability to integrate forensic evidence into the system architecture.
ZF is contributing its extensive expertise in the areas of cybersecurity, IT forensics and neural networks and is working closely with the “Law and Digitalization” research initiative of the Center for Human | Data | Society at the University of Konstanz. The aim is to jointly develop solutions that are both technologically and legally robust. The research program will run for several years and has high social and technological relevance.
Insight into the future: trustworthy AI as an infrastructure principle
Whether in application procedures, telemedicine or networked traffic flows – artificial intelligence will be deeply integrated into everyday processes in the coming years. The results of the FIS program should ensure that these systems remain explainable, resistant to manipulation and, in an emergency, assessable under the rule of law. FIS lays the foundations to ensure that AI systems not only function, but also earn the trust of a free and digital society.
Further information and registration: