> For the complete documentation index, see [llms.txt](https://docs.mediafier.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.mediafier.ai/trust/learning-and-data-boundaries.md).

# Learning & Data Boundaries

Organization learning stays in the organization; platform hardening carries no customer content; network intelligence is not offered today.

> **Your organization's learning stays in your organization.** Mediafier agents get better from the work they do for you; what they learn is yours, and it does not leave your tenant.

***

## Three things that are often blurred

| Concept                   | What it is                                                                         | Does it leave your organization?                                                                                                    |
| ------------------------- | ---------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- |
| **Organization learning** | Decision precedent, validated heuristics, and process memory built from your runs. | **No.** Scoped to your organization; never visible to another tenant.                                                               |
| **Platform hardening**    | Fixes, guards, and safety improvements Mediafier makes to the shared platform.     | Carries **no customer content**. A fix made for one customer's problem class applies to the platform; the customer's data does not. |
| **Network intelligence**  | Cross-organization benchmarks or shared learning.                                  | **Not offered today.** If it is offered, it will be an explicit, opt-in contract — never a default.                                 |

## What the platform learns from your work

* **Decision precedent** — which decisions were made, on what evidence, and which a person approved. Searchable within your organization.
* **Validated heuristics** — proposals from the learning loop that a person reviewed and approved. Nothing proposed becomes active on its own.
* **Process memory** — built only from closed, structured fields of a run (what was asked, what ran, what resulted), never from free-text content or prompt bodies.

## What the platform does not do

* **No fine-tuning.** The learning loop changes the library of governed skills and the directives around them — never model weights. The model is whichever foundation model your team has chosen.
* **No cross-tenant learning.** One organization's runs, decisions, and memories are not an input to another's.
* **No blending of user and organization memory.** What an individual user's session learns is kept separate from the organization's shared learning.
* **No customer content in platform hardening.** The shared platform improves from problem classes, not from your media, prompts, or results.

## How this is enforced

Tenant scoping is enforced at the platform's data layer, not by convention; cross-organization access is tested in CI; and the learning loop's promotion path requires a human decision that is itself recorded. Where the platform cannot show a boundary, it does not claim one.

## What this page is not

* It is not a data-retention schedule and does not present retention as a privacy control; retention terms live in your agreement.
* It is not a benchmark programme. No anonymized cross-customer benchmark is shipped today.

## Where to go next

| If you're…                                | Read next                                                                           |
| ----------------------------------------- | ----------------------------------------------------------------------------------- |
| Reviewing tenancy and identity boundaries | [Multi-Tenant Organization Model](/trust/multi-tenant-organization-model.md)        |
| Understanding how agents get better       | [Decision Tracing & Learning Loop](/platform/decision-tracing-and-learning-loop.md) |
| Supervising what agents are doing         | [Agent Operations](/platform/agent-operations.md)                                   |
| Running an enterprise review              | [Enterprise Readiness](/trust/enterprise-readiness.md)                              |