In the publication, the authors break down the use of AI into a seven-step process, ranging from conception and data acquisition to operational use and maintenance. Their analysis focuses on the intersection between AI technologies and intellectual property (IP) rights.

In the early phase, the primary focus is on clarifying data rights, business models, and the permissibility of text and data mining. During model development, companies must pay particular attention to open-source licenses and the protection of employee know-how. The EU AI Act also sets requirements for transparency and explainability, particularly when evaluating AI results.

During operational use, the focus shifts to IT security, data protection in accordance with the GDPR, and liability issues in the event of malfunctions.

Against the backdrop of the Manufacturing-X project RoX, the authors focus on robotics as a use case. If there are intellectual property (IP) defects in the AI, the operational capability of the entire physical hardware is at risk.

With regard to contract drafting, the publication offers recommendations on due diligence regarding data origin and on risk allocation for third-party models.

Overall, the authors advocate for a proactive, interdisciplinary approach to minimize legal risks early in the development process.