Artificial intelligence is increasingly becoming part of companies operational processes. From production planning, procurement and logistics to quality management and predictive maintenance, new applications also bring growing legal and organizational requirements.
In the podcast “The Transatlantic Perspective”, Daniel Wuhrmann discusses with Vanessa Miller and Raymond J. McVeigh of Foley & Lardner LLP which legal risks companies should consider when using AI in manufacturing and supply chains. The discussion highlights the different regulatory approaches taken by the European Union and the United States.
Different regulatory approaches in the EU and the US
From a European perspective, a key starting point is determining how a specific AI system should be classified and used:
- What role does the company have in relation to the AI system?
- What is the system used for?
- What data does it process?
- Does it have an impact on safety or legally protected rights?
Depending on the application, different regulatory requirements may apply, particularly under the EU AI Act, product safety regulations, data protection law, and cybersecurity requirements.
Contractual arrangements also play a central role. Standard contracts for software or AI solutions are often insufficient for industrial applications. Companies should consider, among other things, performance requirements, responsibilities, data access and usage rights, as well as liability. Documentation requirements and processes for managing changes and updates to AI systems also need to be addressed.
The US takes a different regulatory approach. While companies operating in the EU must particularly consider the requirements of the EU AI Act and other applicable regulatory frameworks, AI-related risks in the United States have so far been addressed to a greater extent through existing legal principles, including contract law, liability, and occupational safety, as well as through risk management and internal corporate governance.
However, this does not mean that companies operating in the US have less need to take action. The podcast recommends, among other things, systematically identifying the AI systems in use and the data they can access, classifying individual use cases according to their risk, and establishing clear internal responsibilities. This should also include rules for human oversight, approvals, and incident management. When procuring AI solutions, performance requirements, data rights, and liability provisions should also be tailored to the specific industrial application.
Conclusion
For both the European and US markets, maintaining a systematic overview of the AI systems in use is essential. Companies should identify which systems and providers they use, what data these applications can access, and what operational consequences could result from incorrect or unintended decisions.
Particular attention should be paid to applications with a high degree of autonomy, for example where an AI system can independently place orders or influence manufacturing processes. For these use cases, companies should clearly define and document responsibilities, the system’s decision-making authority and limitations, as well as appropriate human oversight.
For internationally operating companies, this means that AI governance should not be approached separately for individual markets. Companies using AI systems in Europe and the United States, or offering AI solutions in those markets, need to account for different regulatory frameworks. This includes the classification and risk assessment of AI systems as well as appropriate processes for documentation, allocation of responsibilities, human oversight, and ongoing monitoring and review.
Companies operating in both markets should therefore integrate these requirements into their structures and processes at an early stage. This helps ensure that the use of AI is not only compliant with applicable regulatory requirements but also managed in a transparent, structured, and robust manner.
How do you like this format, and which topics would you like us to cover in future episodes? We welcome your feedback and suggestions.
back