Singulus.ai
About Singulus

AI that nations and organizations can truly own.

Singulus was founded to close a widening gap in the AI landscape: the lack of large language models that nations, public institutions, and privacy-conscious enterprises can deploy without surrendering jurisdiction, data, or values. We're headquartered in Quebec — where some of the world's strictest privacy laws meet abundant low-carbon hydroelectric power. That combination isn't a coincidence. It's our competitive advantage.

Sovereignty

The Sovereignty Charter

Three commitments that govern every model, every deployment, every line of code.

01

Data Sovereignty

Your data never leaves Canadian jurisdiction. No retention. No training on your inputs. Full PIPEDA and provincial law compliance, including Quebec's Law 25. Inputs are processed in volatile memory and discarded immediately on response.

02

Infrastructure Sovereignty

Hardware, cloud, and compute entirely within Canada. We operate our own GPU clusters, with no dependency on foreign hyperscalers. Every component of the stack is auditable and inspectable by qualified customers.

03

Environmental Sovereignty

Training and inference optimized for energy efficiency. We prioritize Quebec hydroelectricity and compression techniques that reduce compute by orders of magnitude. We publish hourly carbon intensity per inference.

Compliance & Certifications

PIPEDAQuebec Law 25SOC 2 Type IIISO 27001 (in progress)Hosted in CanadaZero Data RetentionCarbon Neutral Operations
Our Values

Canadian values, by design

Sovereignty is more than where the servers live. Singulus encodes the values Canada is known for into how our models are trained, governed, and allowed to be used.

Human Rights

Our models are aligned to the rights and freedoms of the Canadian Charter. We red-team for dignity, privacy, and free expression, and decline use cases built for surveillance or repression.

Multiculturalism

Trained for Canada's plural, bilingual reality. English and Quebec French are first-class, with care for the many cultures, dialects, and communities that make up the country.

Equality

We measure fairness across gender, origin, language, and ability — and publish where we fall short. Capable AI should narrow divides, not widen them.

Social Justice

We build for the public good first — healthcare, education, and public institutions over extractive applications. Technology should serve the many, not concentrate power in the few.

Peace

No autonomous weapons, no lethal targeting. Our terms forbid offensive military use. We build AI that informs and de-escalates rather than harms.

Mutual Respect

Consent-first data and transparent governance. We respect the people and communities whose knowledge trains our models — reciprocity, not extraction.

AI Safety

Capable, and safe by design.

As models grow more capable, safety can't be an afterthought. We treat it as a research discipline — engineered into every release, measured against published evaluations, and kept under human control.

Safe by design

Safety lives in the architecture, not bolted on afterward. Guardrails, evaluations, and constraints ship as first-class parts of every model.

Autonomy with guardrails

We build capable autonomous agents — and the safeguards that bound them. Permissioned actions, hard constraints, and human approval on high-stakes steps mean more autonomy never means more harm.

Honest & calibrated

Models are trained to express uncertainty and decline rather than fabricate. We measure and reduce deception — better an honest “I don't know” than a confident error.

Transparent & overseeable

No black boxes you can't inspect. We invest in interpretability and give qualified customers the tooling to audit behaviour and retain oversight.

Team

Our team

A growing team of researchers and engineers in Quebec and across Canada.

We're a small, senior team with backgrounds in large-model training, systems engineering, applied cryptography, and public policy. We hire deliberately — for craft, conviction, and a deep belief that the next decade of AI will be defined by who controls it, not just who builds it.

Contact

For enterprise inquiries, partnerships, government engagements, or press — we'd love to hear from you.

Headquarters

1234 Avenue du Parc, Suite 400
Montreal, Quebec H2X 1Z1, Canada