download dots
AI PromptsCategoriesAI Governance

🧠 Decide Which Models Are Allowed

Put a sensible process around how your organization uses AI, without stopping anyone getting work done. These prompts build the approval queue, the register of what is running, the log of what each tool actually did, and the policy an assistant checks submissions against. Written for the person who has been asked to own this and given no software to do it with.
  • Tested prompts
  • Edit for your use case
  • Works across models
  • Save in Taskade
Decide Which Models Are Allowed

Copy

People pick whichever model is default and I have no view of it. Build me an approved model app. Keep one record per model with the provider, where it runs and where data is processed, whether inputs may be used for training, the data classifications it may be used with, the use cases it is approved for and any explicitly prohibited, its cost characteristics, the date approved, the version or date the approval applies to, and the review date. Make clear that an approval attaches to a specific version, since vendors update models frequently. Show a simple view people can consult before starting work: given what I am doing and the data involved, which models may I use. Add an AI agent that answers that question in plain language from the records, explains why a particular model is not suitable for a given case, and suggests the closest approved alternative. Automate two things: a review trigger when a vendor announces a material model change, and a scheduled periodic review of the whole list. Two logins: a user who consults the list, and a governance owner who maintains approvals.

Run this prompt with AI
STATUS · LIVE
v1

01 / Spark - intelligence awakens

Decide Which Models Are Allowed

Maintain approved models with permitted uses, data restrictions, hosting location, and review triggers on vendor changes.

Start with AI
// the spark catches · intelligence awakens · the answer unfolds

Model choice is usually made by whoever set the tool up. This records which models are approved for what, so the decision is deliberate and reviewable.

What's Included

  • Approved models with permitted uses and clear restrictions.
  • Where each model runs which is often the deciding factor.
  • A review when a model changes because vendors update constantly.
  • An assistant that suggests the appropriate approved model.

How To Use

  1. Approve by use, not by model alone. The same model is fine for one task and not another.
  2. Record where the model runs and what happens to the data.
  3. Review when a vendor updates. An approval attaches to a version.
  4. Keep the list short enough that people actually consult it.

Link it to the policy register, and enforce it in the governance office.

Discover more

More Governance and Oversight Prompts

View all Governance and Oversight Prompts