Model Policy
Model policy controls which AI models are available for governed work in Clear Ideas. It helps administrators align model access with privacy posture, task requirements, plan availability, and organization governance requirements.
Model policy applies to model-aware experiences such as AI Chat and Agents. Site settings and user choices can narrow behavior further, but they cannot exceed organization policy.
Policy Concepts
Model policy can support several approval patterns:
- All approved models - users can choose from the available model catalog.
- Private Models only - users can choose private execution model lanes where enabled.
- Provider Models only - users can choose approved external provider models.
- Private and Provider Models - users can choose either model lane.
- Latest aliases - Agents can follow Clear Ideas-managed aliases for future scheduled runs.
- Specific models - administrators approve named concrete model versions for stricter repeatability.
- Deprecated models already in policy - older selections remain visible so administrators can remove or replace them deliberately.
Private Models vs Provider Models
Use Private Models when private execution, cost predictability, and repeatable governed execution matter most.
Use Provider Models when a task benefits from premium reasoning, specialized capability, broad multimodal support, or a particular provider model approved by the organization.
Most organizations should allow a deliberate mix rather than treating either lane as universally best.
Latest Aliases and Concrete Models
Agents can run later, so model policy may allow latest aliases. A latest alias lets future scheduled runs follow the current Clear Ideas production recommendation.
Choose a concrete model version when repeatability matters more than automatic upgrades, such as benchmark comparisons, regulated review, or process validation.
Deprecated Models
Deprecated models are not automatically replaced by successors. If an Agent or run references a deprecated or no-longer-permitted model, update it to an active permitted model before running it.
Deprecated model choices may remain visible in policy or older Agent definitions so administrators can see and clean up historical configuration intentionally.
Where Policy Applies
Model policy can affect:
- model choices in AI Chat
- Agent default model and reasoning settings
- step-level Agent model overrides
- scheduled Agent runs
- private execution availability
- generated file helpers when model skills are required
Review Guidance
Review model policy when:
- enabling Private Models
- changing provider-model access
- publishing scheduled Agents
- using generated files in recurring runs
- replacing deprecated models
- responding to privacy, compliance, or customer requirements