GPAI vs High-Risk in the Plant: What Applies to Your Use Case?

The EU AI Act introduces a new category: General-Purpose AI (GPAI). For manufacturers, understanding whether an AI model counts as GPAI or high-risk determines documentation scope, liability, and vendor responsibilities.

What Is GPAI?

General-Purpose AI refers to models that can be adapted for many applications, such as large vision models or pre-trained foundation models. They’re usually provided by upstream suppliers or cloud vendors.

When GPAI Becomes High-Risk

• If your factory fine-tunes a GPAI model for a regulated task (e.g., defect rejection), it inherits high-risk obligations.
• If you use a GPAI service without modification (e.g., for analytics), the supplier bears primary compliance duties.

How to Classify Your Use Case

1. Identify model origin (internal, vendor, or open source).
2. Define intended use and whether it affects compliance or safety.
3. Check if fine-tuning or retraining changes the model’s purpose.

Responsibility Matrix

Case Example: Plastic Injection Plant

Using a cloud foundation model fine-tuned for cavity defects, the plant became a co-provider under the AI Act. A shared conformity statement was filed with the vendor.

Related Articles

• EU AI Act for Manufacturing: Is Your Vision System ‘High-Risk’?
• AI Governance for Factories: Policies, Logs, and Human Oversight
• Timeline to Compliance: A Factory’s 12-Month Plan

Conclusion

Determining whether your system is GPAI or high-risk is essential for compliance. Manufacturers must clarify roles early — ideally in supplier contracts — to avoid regulatory ambiguity later.