Adaptive policies for secure and trustworthy AI agents
Artificial intelligence is becoming increasingly important in enterprises, particularly in the areas of automation, decision support, and data-driven processes. This raises the central question of how AI agents can act reliably, transparently, and in compliance with applicable rules and regulations. The present project addresses this challenge by developing an adaptive framework that automatically generates, verifies, and enforces policies.
The project develops an adaptive framework that enables AI agents to operate securely, transparently, and in a compliant manner. The objective is to automatically generate, validate, and enforce policies so that agent behavior consistently aligns with user intent, organizational requirements, and legal regulations.
The research project delivers a prototype framework that enables organizations to deploy trustworthy, auditable, and regulation-compliant AI agents. By combining fortiss’ expertise with the domain-specific knowledge of the project partners, this initiative strengthens regional innovation capacity in Bavaria and fosters collaboration between research and industry.
We are seeking an SME partner with a highly autonomous AI agent use case. The AI agent should be used or planned in a critical domain where failures such as data leakage or misuse could have serious impact. Deployment is not required, as the project aims to develop a proof of concept.
• provide a use case involving highly autonomous AI agents.
• operate in or target critical, mission-critical, or regulated environments (e.g., industrial systems, healthcare, mobility, energy, critical infrastructure, finance, public services).
• have sufficient technical and organizational capacity to collaborate in a research project, including access to relevant systems, data, or expertise.
The partner will get early access to the developed framework and gain expertise by applying it to their use cases. Additional benefits include collaboration with fortiss on cutting-edge research, influence on emerging AI standards, and opportunities for joint publications, demonstrators, and potential follow-up funding.
