GPAI Dashboard

The GPAI Dashboard shows General-Purpose AI models in the current space and highlights whether each model has the metadata Qarion needs for EU AI Act Article 52-55 obligations.

Use this page when you need to find GPAI models with missing public summaries, downstream-use documentation, policy links, systemic-risk tier data, or open-source license context.

Where The Data Comes From

A model appears on /gpai-models when its AI model details have is_gpai set to true. The dashboard reads model detail fields from catalog products; it does not create or edit GPAI records directly.

To inspect or remediate an item:

1. Open /gpai-models in the target space.
2. Review the missing-field badges for each model.
3. Use the row action to open the product.
4. Review Model Details and the GPAI obligations section.
5. Update the model metadata through the catalog or model repository workflow used by your team.

Summary Cards

Card	Meaning
Total Models	Count of products in the current space with is_gpai enabled.
Fully Compliant	Count of GPAI models with compliance_percentage equal to 100.
Systemic Risk	Count of GPAI models treated as systemic risk.
Open Source	Count of GPAI models with an open-source license value.

Table Columns

Column	Meaning
Model	Catalog product name.
Type	Catalog product type.
Tier	Effective GPAI tier: standard or systemic risk.
Provider Role	Organization role, such as provider, deployer, or distributor.
Compliance	Metadata completeness percentage.
Missing Fields	Required fields that are currently empty.

Compliance Calculation

Standard GPAI models require these fields:

• training_data_summary
• copyright_policy_url
• acceptable_use_policy_url
• downstream_obligations_doc

Systemic-risk models require those fields plus:

• gpai_tier

Open-source models are identified by open_source_license. When an open-source license is present, Qarion does not require copyright_policy_url or acceptable_use_policy_url for the GPAI dashboard score.

The compliance percentage is the rounded share of required fields that are filled. Empty strings and empty lists count as missing.

Systemic-Risk Detection

Qarion treats a GPAI model as systemic risk when either condition is true:

• gpai_tier is systemic_risk.
• training_compute_flops is greater than or equal to 1e25.

The Model Details tab also shows the training compute value and flags models that exceed the 1e25 FLOPs threshold.

Source Fields

Field	Use
is_gpai	Includes the model in the GPAI dashboard.
gpai_tier	Records standard or systemic_risk.
gpai_provider_role	Records the organization's relationship to the model.
training_data_summary	Public summary of training content.
training_compute_flops	Estimated training compute used for systemic-risk detection.
copyright_policy_url	Copyright compliance policy URL.
acceptable_use_policy_url	Acceptable-use policy for downstream providers.
downstream_obligations_doc	Integrator or downstream obligations document.
open_source_license	Open-source license identifier and scoring exception signal.

Troubleshooting

Symptom	What to check
No models appear	Confirm the product has AI model details and is_gpai is true.
A model is marked systemic risk unexpectedly	Check whether training_compute_flops is at least 1e25 or gpai_tier is systemic_risk.
Policy fields are not counted as missing	Check whether open_source_license is populated.
Compliance percentage is lower than expected	Open the missing-field badges and fill the named source fields on the model record.
The dashboard does not update after edits	Refresh the product and dashboard, and confirm you edited the model in the same space.

Developer Reference

See the GPAI Compliance API for the product compliance endpoint, space listing endpoint, response fields, pagination headers, and scoring rules.