Sovereign AI
What is sovereign AI?
Sovereign AI is a nation's capacity to develop, deploy, and govern artificial intelligence on infrastructure and under laws it controls. For Canada, it means frontier models built and served domestically, with data protected under Canadian privacy law, so that critical intelligence does not depend on foreign providers.
Dipole ML Inc. is a Toronto-based Canadian sovereign frontier AI lab that builds the Raven model family, DCode, and the DipoleML API. This page explains what sovereign AI means, why it matters for Canada, and how we put it into practice.
Why does sovereign AI matter for Canada?
When a country's businesses, researchers, and public institutions run their most sensitive workloads on foreign frontier models, three things leave with them: the data, the provenance, and the control. Requests fall under another jurisdiction's laws, routing passes through third-party brokers, and access can change with another market's pricing or politics.
Canadian organizations already operate under strict statutory privacy obligations, including PIPEDA. Sovereign AI closes the gap between those obligations and where the intelligence actually runs: it makes the legal promise and the technical reality match.
What does sovereign AI look like in practice?
Sovereignty is not a badge on a data center; it is a property of the whole serving path. In practice it means the organization operating the models runs its own gateway and model serving infrastructure, so every request has direct end-to-end provenance with no third-party middleman brokers between the user and the model.
It also means accountability. A sovereign provider answers to its users rather than to an advertising business model, and its infrastructure sits under the jurisdiction and the privacy statutes of its home country.
How does Dipole ML approach sovereign AI?
Dipole ML was built as a sovereign frontier lab from first principles. Every architectural choice below is already how the products ship today:
Built and served from Canada
Raven models and DCode are engineered, served, and billed from Canada, under Canadian jurisdiction, with no third-party middleman brokers.
Our own gateway and serving infrastructure
We operate our own compute gateway and model serving infrastructure, giving every request direct end-to-end provenance.
Zero data retention
Inference requests are processed ephemerally in memory and discarded immediately after generation. We never log your content and never use customer data to train models.
DEK-encrypted chat history
Saved conversations in the Raven app are encrypted at rest under individual Data Encryption Keys. Engineering staff and administrators have zero visibility into transcripts.
Inspectable reasoning
Raven models stream their reasoning in plain view, so you can verify how an answer formed. Nothing happens in an opaque black box, and every agentic action in DCode requires your approval.
Who is sovereign AI for?
Anyone whose data, decisions, or obligations stay in Canada: enterprises and regulated institutions that need direct provenance, healthcare organizations, public sector bodies, and developers who want frontier performance without sending their work abroad. The same models serve a student and a ministry, because access was designed in from the start.
For organizations that need dedicated Canadian compute instances, custom SLAs, or invoicing, tailored sovereign deployments are available through our support team.
Where can I try sovereign AI today?
Every surface is live now: chat with Raven in the Raven app, build with DCode, or call the DipoleML API. Free tiers start at zero, and the pricing page lists every plan in USD.