ProductivIA is designed to give access to generative AI without giving up data governance. This page explains, first in plain language and then in more detail, how the platform frames hosting, external providers, data, client applications, the OS, security and responsible use.
AI platform compliance is not only about where servers are located. It rests on several layers: applicable laws, each organization's internal rules, the nature of the data, the providers requested, infrastructure security, permissions, retention, auditability and the way users are guided.
ProductivIA takes a pragmatic approach: clearly explain what is already in place, what depends on the selected configuration, what must be validated in a client context, and what is part of an ongoing process. This page does not replace a legal opinion, a privacy impact assessment or an external audit, but it provides a serious due diligence basis for institutions, businesses, the education sector and individuals.
ProductivIA works like a complete working environment: desktop, windows, taskbar, notifications, files, apps and shared services. This architecture centralizes security and governance rules at the system level instead of leaving each application to manage its own practices alone.
Each organization can have a distinct silo. A silo contains its users, sessions, contacts, application data, settings, authorized apps, keys, rules and permissions. The separation is physical in how data is organized, not only logical in a shared database.
ProductivIA distinguishes several data categories: user account, settings, entered content, imported files, generated documents, conversation history, AI consumption, technical logs, administrative data and app-specific data. These data categories do not all have the same sensitivity or retention period.
The platform can mobilize several model families: sovereign models, local models, models hosted in Canada when available, and specialized external providers for certain tasks. This capability is useful for quality, cost, resilience and performance, but it must be governed.
By default, the intended approach is simple: use internal or sovereign capabilities first for common tasks, then route to an external provider only if the task justifies it, if the context is appropriate and if the organization's rules allow it. When an external provider is used, ProductivIA requires a clear contractual commitment: no retention of transmitted content beyond processing the request, no model training on that data and no reuse for other purposes.
An important clarification: ProductivIA does not claim that every request always remains in the same place in every mode. If an organization activates an external provider, the request needed for processing may be transmitted to that provider according to the selected settings and applicable terms. In return, the no-retention commitment is contractually documented, audited and renewable. Compliance therefore rests on a combination of configuration, data qualification, consent when required, contractual framing and logging.
Fabrique makes it possible to generate applications from a natural-language description. This power requires a clear rule: a generated application should not automatically become an official tool for an entire organization without validation. ProductivIA therefore provides a controlled lifecycle.
For educational institutions, this cycle helps distinguish pedagogical experimentation, classroom tools, administrative tools and institutional tools. For businesses, it limits the spread of uncontrolled small apps. For institutions, it supports accountability.
ProductivIA can be used in a modern browser, in a dedicated server environment, or in a more sovereign chain that includes Boréal OS. The required level depends on the type of data, legal framework, user profile and organizational requirements.
Choosing a mode is not ideological. It should follow risk: public or common educational data, internal business data, personal information, sensitive data, strategic information, institutional records or content subject to specific rules.
ProductivIA emphasizes logging, permissions and observability. Administrators must be able to understand who accesses what, which models are used, which apps are active, which costs are incurred and which incidents or errors must be handled.
Generative AI models can produce errors, omissions, approximations, bias or plausible but false content. ProductivIA is therefore designed as an assisted production environment, not as an automatic authority.
The recommended rules are:
ProductivIA's operational priorities are: