gpai-code-of-practice

A GPAI model is classified as systemic risk if its training compute exceeds 10^25 FLOPs or if it receives a Commission designation. This classification triggers Tier 2 safety-and-security planning and additional systemic risk obligations.

The EU AI Act open-source exemption under Article 53(2) can be evaluated through a specific workflow. If the exemption does not apply, downstream consequences include Tier 1 obligations such as training data disclosure, copyright compliance, and transparency requirements.

DACH-specific enforcement timelines for the EU AI Act GPAI requirements outline the compliance schedules for Germany, Austria, and Switzerland. These regional considerations must be factored into your provider compliance planning and systemic risk mitigation.

Training data disclosure is required for GPAI providers under Tier 1 transparency obligations of the EU Code of Practice. You must complete and submit a Model Documentation Form detailing your training data unless a valid open-source exemption applies.

A provider determination workflow classifies GPAI obligations by mapping your model to Tier 1 requirements like transparency and copyright, and Tier 2 systemic risk requirements. This process evaluates compute thresholds and provider status to define mandatory actions.