Making the right decision with a thorough analysis
AI agents are taking on increasingly complex tasks: they make decisions, process sensitive data, and automate entire workflows. For small and medium-sized enterprises (SMEs), however, this entails significant challenges:
- Loss of control: How can it be guaranteed that an AI agent only performs actions explicitly authorized by the user?
- Vulnerabilities: AI systems can be manipulated through targeted inputs (known as jailbreak or prompt injection attacks) to disclose trade secrets or bypass established rules.
- Bureaucracy: Complying with strict regulations, such as the European AI Act, is barely manageable for SMEs without costly external expertise.
Our Approach
Together with you, we are developing a security framework that makes AI agents completely reliable and verifiable. The project is based on four pillars:
- Reliable Identity: Every action performed by the AI is strictly linked to the user’s identity and actual intent.
- Real-Time Protection: Attacks and manipulation attempts are blocked before the agent executes the action.
- Automated Compliance: Legal requirements (e.g., the EU AI Act) and internal company policies are automatically translated into concrete safeguards for the AI.
- Verifiable Evidence: The system automatically generates digital records (assurance cases) to demonstrate to customers and regulatory authorities that your AI operates securely and in compliance with the law.
Project Goal
A practical prototype that enables SMEs to deploy AI agents in their daily operations without security or compliance concerns. Furthermore, the joint development of secure and regulatory-compliant agent systems provides the opportunity to open up new, future-oriented fields of innovation and business for your company.
Your use case as a starting point for innovation
We are looking for a partner to contribute a concrete industrial or operational use case for AI agents. Relevant scenarios include, for example:
- Process Automation: Autonomous workflows within internal systems (e.g., ERP, CRM, HR management).
- Knowledge and Data Management: Intelligent assistant systems for the automated analysis, aggregation, and processing of complex corporate data.
- Market and Trend Analysis: Agent-based systems for continuous, automated information retrieval and evaluation.
Your role in the project
- Requirements definition: You provide the industrial use case and define the operational and regulatory requirements based on real-world practice.
- Validation: You support the testing and evaluation of the prototype under real-world conditions in your system environment.
- Transfer: You participate in the application of the results and present the findings to your industrial network.
Your benefits at a glance
- Technological edge: Exclusive, early access to a highly innovative framework for securing your AI applications.
- Legal certainty: Direct expertise and technical solutions to ensure compliance with the EU AI Act within your own organization.
- Knowledge transfer: Direct collaboration with fortiss researchers, co-publications, and opportunities for future consortium grants.
- Funding: Joint application under the “KMU Innovativ” funding program. Once funding is approved, funds will be allocated for the research project.
Partner profile
We are looking for an SME in the fields of artificial intelligence, software development, or cybersecurity that wishes to implement autonomous agent structures in business use cases. Prior experience in access control, compliance, or large language models (LLMs) is a plus but not required.