liveBook · Manning

General-purpose AI provider

As explained in AI Governance

A general-purpose AI provider is an entity that develops or places a general-purpose AI model on the market under its name or trademark.

A general-purpose AI provider develops or places a general-purpose AI model on the market under its name or trademark. Its role is distinct from that of a downstream builder or deployer: provider documentation supports their compliance work but does not replace their own risk assessment, oversight, or monitoring. section 7.2.4

When integrating a model, obtain evidence that downstream teams can use, including:

  • Documentation of capabilities and known limitations, foreseeable misuse scenarios, integration guidance, and interpretability information. section 7.2.4
  • A copyright policy for training data, a public summary of training data, and enough information to support downstream obligations. section 7.2.4
  • Contractual commitments to provide documentation before signing; update the evidence pack when the model or provider changes. section 7.2.4

A vendor API is not a compliance handoff

Using the OpenAI API does not transfer the integrator’s risk assessment or monitoring duties. If a provider update breaks accuracy benchmarks, monitoring that change remains the integrator’s responsibility. section 7.2.4

Figure 7.10 How to obtain legally required information from a GPAI provider.

Use the figure to navigate the process for obtaining legally required information from a provider. section 7.2.4

How AI Governance covers it

How deep it goes
Specifies documentation contents, contract demands, complaint routes, and update duties, while deferring the full checklist.
Examples
Uses the OpenAI API, provider updates that break accuracy benchmarks, and switching from OpenAI to Anthropic or Llama.
What you can do after
Evaluate provider evidence by checking documentation, requiring it contractually, and updating records after provider changes.
How it connects
Connects model providers to downstream builders, deployers, the EU AI Act, and the GPAI Code of Practice.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.