DFKI Guidelines for Responsible AI Research in the Defense and Security Sectors
The German Research Center for Artificial Intelligence (DFKI) develops AI technologies grounded in European values, scientific excellence, and social responsibility. Even in the defense and security sectors, our research is guided by clear ethical, legal, and democratic principles. The goal is to develop AI systems that support people, promote security, and help protect society, democracy, and critical infrastructure.
Our Core Principles
Artificial intelligence research and development at DFKI is guided by five overarching principles:
• Human-centered approach: AI should support people, enhance their capabilities, and be operated under human supervision.
• European Responsibility with a Global Perspective: We develop AI solutions that deliver societal benefits and strengthen trustworthy European AI development.
• Ethics by Design: Ethical considerations are integrated throughout a system’s lifecycle—from conception to decommissioning. Transparency, fairness, non-discrimination, and explainability are central requirements.
• Environmental Responsibility: Environmental impacts are considered, and sustainability goals are incorporated into the development and operation of AI systems.
• Social Dialogue: DFKI promotes open discussion of the opportunities, risks, and impacts of AI and supports an informed public debate.
AI Research and Development at DFKI
DFKI develops AI systems that help people and organizations achieve their goals. Safety, reliability, and human oversight are central to this effort. AI is not seen as a substitute for human responsibility but as a tool that supports decisions and processes.
Our research always considers AI within the context of the entire socio-technical system comprising people, organizations, and technology. In safety-critical application areas, effective collaboration between humans and machines is a key prerequisite for responsible innovation.
Guidelines for Defense and Security-Related Research
Additional specific requirements apply to projects in the defense and security sectors.
1. Protection of Human Life and Human Decision-Making Responsibility
The DFKI does not engage in research or development of lethal autonomous weapon systems. Decisions about the use of lethal force must not be delegated to autonomous systems.
2. Responsible Use of AI
Defense-related research projects must promote lawful, transparent, and controllable AI systems. Responsibilities must be clearly defined, human oversight must be maintained, and the risk of discriminatory or biased decisions must be minimized.
3. Ethical Review of All Relevant Projects
Projects with military, intelligence, or comparable relevance undergo a structured evaluation. This evaluation assesses potential impacts on human rights, societal security, dual-use risks, and compatibility with DFKI’s values and principles.
4. Exclusion of Certain Applications
DFKI does not develop AI systems for public mass surveillance, mass biometric identification, social scoring, individual or societal manipulation, or targeted discrimination.
5. Focus on Responsible AI for Protection and Defense
DFKI's defense and security research focuses particularly on application areas with societal benefits, including:
• Defensive cybersecurity: Protection of critical infrastructure against cyberattacks, intrusion detection, and anomaly detection.
• Logistics and supply: Optimization of supply and rescue chains in crisis situations and resource planning.
• Search and rescue systems: Autonomous or semi-autonomous systems for search and rescue operations in hazardous environments.
• Mine clearance: AI-supported systems for detecting and removing mines and unexploded ordnance.
• Medical care: Telemedicine, AI-supported diagnostics, and treatment assistance in crisis areas.
• Infrastructure protection: Monitoring and securing critical supply infrastructure without resorting to mass surveillance.
• Early warning systems: Systems that predict and detect threats to public safety at an early stage.
This list is not exhaustive. Other areas of application may be classified as excluded or permissible after review through the ethical assessment process.
Our Commitment
DFKI combines academic freedom with social responsibility. We are convinced that AI can build lasting trust in the defense and security sectors only if its development is transparent, traceable, and consistently aligned with human rights, democratic values, and the protection of human life. These guidelines for responsible research provide a framework that sets clear boundaries while promoting innovation.