model_origin
Level: activity
Values
commercial- The model comes from a commercial provider (for example OpenAI, Microsoft, Google or Anthropic) and is used through its product, licence or programming interface.
open_weight- The model's weights are openly released (for example Llama, Mistral or Falcon), whether the institution or a provider runs it.
government_developed- The model was developed by or for a government (for example France's Albert or Germany's OpenGPT-X), including by the institution's own team.
Missing answers
unknown- The page mentions the model behind the tool but does not let it be settled.
not_documented- The page is silent.
build_arrangement
Level: activity
Values
off_the_shelf_product- The institution uses a ready-made product as supplied (for example a commercial chat assistant or an office-suite assistant), set up but not built on.
custom_application_on_external_model- The institution, or a contractor for it, built its own application (a chatbot, assistant or workflow) on a model it did not train.
model_trained_or_fine_tuned_by_institution- The institution, or a contractor for it, trained or fine-tuned the model itself. Internal use alone does not show this.
Missing answers
unknown- The page mentions how the tool was built but does not let it be settled.
not_documented- The page is silent.
interface
Level: activity
Values
conversational- Users type or speak to the tool in a chat or question-and-answer exchange, as with a chatbot or assistant window. A tool the page calls a chatbot or assistant counts.
backend_or_api- The tool runs behind the scenes, or other software calls it through a programming interface; users do not talk to it directly.
other_interface- Users reach the tool in another way than a chat exchange or a back-end connection.
function_drafting_writing
Level: activity
Values
yes- The page states that the tool is used for drafting or writing text, such as memos, letters, press releases or social media posts.
no- The page does not state that the tool is used for drafting or writing text, such as memos, letters, press releases or social media posts.
function_summarization
Level: activity
Values
yes- The page states that the tool is used for summarising documents, reports or meetings.
no- The page does not state that the tool is used for summarising documents, reports or meetings.
function_translation
Level: activity
Values
yes- The page states that the tool is used for translating text or speech between languages.
no- The page does not state that the tool is used for translating text or speech between languages.
function_search_retrieval
Level: activity
Values
yes- The page states that the tool is used for finding information, such as answering questions from documents or searching the institution's own records.
no- The page does not state that the tool is used for finding information, such as answering questions from documents or searching the institution's own records.
function_code
Level: activity
Values
yes- The page states that the tool is used for writing, reviewing or fixing software code.
no- The page does not state that the tool is used for writing, reviewing or fixing software code.
function_image_audio_video
Level: activity
Values
yes- The page states that the tool is used for creating or editing images, audio or video.
no- The page does not state that the tool is used for creating or editing images, audio or video.
function_extraction_classification
Level: activity
Values
yes- The page states that the tool is used for pulling information out of documents or forms, or sorting and classifying them.
no- The page does not state that the tool is used for pulling information out of documents or forms, or sorting and classifying them.
function_analysis_recommendation
Level: activity
Values
yes- The page states that the tool is used for analysing information or giving recommendations or scores to support human decisions, such as case triage or eligibility checks.
no- The page does not state that the tool is used for analysing information or giving recommendations or scores to support human decisions, such as case triage or eligibility checks.
deployment_mode
Level: activity
Values
integrated- The GenAI feature works inside software or a service channel the institution already used (office suite, case system, existing website service).
standalone- A separate application or site created for the GenAI tool.
Missing answers
unknown- The page mentions whether the tool is separate or built into existing software but does not let it be settled.
not_documented- The page is silent.
target_users
Level: activity
Values
internal_staff- Only the institution's own staff use the tool.
public- People outside the institution use the tool: members of the public, residents or businesses.
both- Both the institution's staff and people outside it use the tool. On long pages, one part naming staff and another the public also gives this value.
Missing answers
unknown- The page mentions who uses the tool but does not let it be settled.
not_documented- The page is silent.