CanGPT: Canada’s Public AI Answer to ChatGPT for Digital Sovereignty

The blue light of my screen cast a soft glow on my living room, the late evening quiet punctuated only by the subtle whir of my laptop.

I was reading another article about ChatGPT, Google Gemini, and the seemingly boundless commercial expansion of generative AI.

It was impressive, certainly, but a familiar unease settled in.

Like many Canadians, I found myself wondering: are we simply to be consumers of these powerful new technologies, shaped by the private interests that build them?

Or is there another path, one rooted in public good, echoing Canadas own rich history of public service?

This reflection is not merely academic; it is a vital conversation about the soul of our digital future, prompting us to consider if AI can truly serve all of us, not just shareholders.

In short: CanGPT is a proposed national, public-service AI model for Canada, taking inspiration from public broadcasters.

It aims to prioritize public values, democratic control, and energy efficiency, offering a distinct alternative to commercially driven generative AI tools like ChatGPT.

Why This Matters Now: The Erosion of the Digital Commons

The digital landscape is being rapidly reshaped by generative artificial intelligence.

In Canada, much of the discourse has focused on commercial innovation, celebrating the technical prowess of tools like ChatGPT and Google Gemini.

This focus, however, often overlooks a critical truth: these innovations depend heavily on access to global cultural knowledge—the very internet, treated as a knowledge commons.

The paradox is stark.

AI, in its current commercial form, would have been impossible without public data, much of which was freely consumed without directly contributing back to the public system.

This practice creates an imbalance, where the digital commons are harvested for private profit, raising significant questions about fairness, access, and the broader public interest.

As AI becomes more deeply embedded in our lives, from content creation to critical decision-making, the imperative to ensure its development aligns with public values, rather than solely commercial incentives, grows ever more urgent.

The Core Problem: AIs Private Gatekeepers

Imagine the early days of automated translation.

In the 1980s, Canadian parliamentary transcripts, filled with multilingual material, were anonymously sent to IBM on a tape reel.

This public resource helped train early translation algorithms, becoming a foundational element for what would later evolve into sophisticated AI tools.

This historical link reveals a profound truth: much of AIs power is built on collective knowledge, on public data.

The core problem today is that while AI relies on this shared commons, the guardrails governing its use—especially for addressing harms like deepfake pornography and technology-assisted violence—are predominantly set privately by tech companies.

Some platforms opt for minimal moderation, while others, like OpenAI, might ban certain uses by politicians.

These private decisions carry immense political implications, shaping content moderation and influencing social media governance without direct democratic input.

The counterintuitive insight here is that by allowing private entities to define these crucial boundaries, we inadvertently cede democratic control over a technology that impacts every facet of public life.

What the Research Really Says: A Public Path for Canadas AI Strategy

The idea of a national, public AI model, often dubbed CanGPT, emerges from a growing concern that commercial AI development, while impressive, fundamentally operates on assumptions that may not align with public good.

Research into this concept reveals several critical findings for Canada.

First, commercial AI heavily relies on user-generated and public data, frequently without direct contribution back to public systems.

The so-what is clear: public knowledge is privatized, benefitting corporations without equitable return to the public.

The implication for Canada is that a public-service AI model could ensure the vast amount of public data used for training AI is leveraged explicitly for public good and democratic values, rather than exclusively commercial profit.

Second, existing guardrails for generative AI harms, such as deepfake pornography and technology-assisted violence, are currently set privately by tech companies.

The so-what is that decisions with profound political and social implications are made without democratic oversight.

The implication is that a publicly governed AI model like CanGPT would allow Canadians to democratically debate and define ethical boundaries and content moderation policies for AI through public institutions, thereby contrasting private decision-making with public values.

Third, Canadas federal government has made significant investments, including billions, in a costly AI Sovereign Compute Strategy.

The so-what is that this infrastructure-heavy approach might be ineffective, potentially benefiting American firms and dismantling Canadas capacity to build public-interest AI, while also carrying a large environmental impact.

The implication is that a public-good framework for AI could advocate for frugal, energy-efficient models running on smaller, local machines, prioritizing targeted tasks with a lower environmental footprint, offering a less risky future if the AI bubble bursts.

CanGPT: A Canadian Model for Public AI

The vision of CanGPT is not entirely new; a growing number of countries, including Switzerland, Sweden, and the Netherlands, are already experimenting with national or publicly governed AI models to create public AI services.

Canadas own federal service has an internal tool called CanChat, an alternative to ChatGPT, but it is not publicly accessible.

The model for CanGPT draws heavily from Canadas enduring success with public service media, namely the CBC and Radio-Canada.

Just as public broadcasters emerged to ensure new communication technologies served democratic needs when radio and television first appeared, a similar approach could work for AI.

Instead of allowing companies to dictate the future of AI, Canadian Parliament could sponsor its own AI model, potentially expanding the mandate of an organization like the CBC to deliver better access to AI.

Such a public model could harness vast resources: materials in the public domain, government datasets, and publicly licensed cultural resources.

Crucially, CBC/Radio-Canada possesses an enormous, multilingual archive of audio, video, and text spanning decades.

If treated as a public good, this corpus could become a foundational dataset for a Canadian public-service AI.

CanGPT could be an open-source system, available as an online service or a locally run application, thereby providing public access and anchoring a broader national AI strategy rooted in public values rather than commercial incentives.

This approach to digital sovereignty ensures national interests are prioritized.

Setting Democratic Boundaries for AI

The development of CanGPT would force a much-needed national conversation about what AI should and should not be able to do.

This is a crucial aspect of AI governance.

Generative AI is already implicated in serious harms, including deepfake pornography and various forms of technology-assisted violence.

Currently, the ethical boundaries and guardrails for these harms are typically set by private tech companies.

Their platforms might have minimal moderation, or, like OpenAI, they may ban politicians and lobbyists from using ChatGPT for official campaign business.

These are profound political decisions that shape content moderation and social media governance, yet they are made without direct public input.

A publicly governed AI model like CanGPT could fundamentally change this narrative.

It would empower Canadians to debate and define these acceptable-use policies and content moderation rules through democratic institutions, rather than leaving such critical decisions to technology firms.

This shift would align AI ethics more closely with Canadian values.

The Environmental and Economic Impact of Public AI vs. Commercial AI

Canadas current approach to AI, heavily reliant on an infrastructure-heavy AI Sovereign Compute Strategy, has seen billions invested by the federal government.

This strategy, despite growing concerns about an AI bubble, risks being ineffective, potentially benefiting American firms and dismantling Canadas capacity to build public-interest AI.

Moreover, Canadas AI agenda has a significant environmental impact, reflecting massive data centre investments.

A public-good framework for AI, as envisioned with CanGPT, offers a stark contrast.

It could encourage the development of frugal, energy-efficient models that run on smaller, local machines.

These models would prioritize targeted tasks rather than the massive, multi-billion parameter models like ChatGPT, which have a substantial environmental footprint.

A smaller, public model could significantly contribute to a lower environmental impact, offering a less risky and more sustainable future if the AI bubble bursts.

This approach aligns with digital transformation efforts that prioritize sustainability.

A Playbook for National Public AI

• First, model after public service media.

  Draw inspiration from successful public broadcasters like CBC/Radio-Canada to understand how new technologies can serve democratic needs and public values.

  This provides a foundational framework for a public utility approach to AI.

• Second, leverage national data assets.

  Utilize existing public domain materials, government datasets, and publicly licensed cultural resources as foundational datasets.

  CBC/Radio-Canadas extensive multilingual archive, for instance, offers a rich corpus for training a Canadian public-service AI.

• Third, establish democratic governance structures.

  Create mechanisms that allow citizens to democratically debate and define AIs ethical boundaries, content moderation policies, and acceptable-use policies through public institutions, rather than relying on private tech firms.

• Fourth, prioritize open-source development.

  A national public model should be developed as an open-source system, available either as an online service or a locally run application.

  This ensures public access and fosters transparency.

• Fifth, focus on energy efficiency and targeted tasks.

  Advocate for frugal, energy-efficient AI models that run on smaller, local machines.

  This contrasts with massive, resource-intensive commercial models, reducing environmental footprint and financial risk.

• Finally, anchor a broader national AI strategy.

  Position the public AI model to anchor a wider national AI strategy, one rooted in public values rather than solely commercial incentives.

  This ensures AI development serves national interests and ethical standards.

Risks, Trade-offs, and Ethical Considerations

Building a national public AI model like CanGPT would not be simple.

Significant questions remain about how to fund it, how to ensure it is regularly updated, and how to maintain competitive performance compared with rapidly evolving commercial AI offerings.

These are practical trade-offs that demand careful consideration and substantial long-term commitment.

The ethical stakes are particularly high.

Beyond content moderation, which CanGPT could address through normative principles, a public AI initiative would necessitate an ongoing, transparent dialogue about AIs social purpose.

This includes discussions on potential biases embedded in training data and the accountability mechanisms for AIs outputs.

Ensuring a diverse and representative input into its development is crucial for maintaining public trust and avoiding unintended consequences.

The risk of political influence, even in a public model, must also be acknowledged and mitigated through robust independent oversight.

The goal is to create an AI that serves the public without becoming a tool for any single political agenda.

Tools, Metrics, and Cadence for Public AI Success

Developing and maintaining a public AI like CanGPT would require a specific set of tools and a clear operational cadence.

Essential Tools

• Sovereign Cloud Platforms specifically designed public, private, and partner cloud offerings that adhere to data residency and regulatory compliance.
• AI Development and Deployment Platforms, such as integrated services like Azure AI and Microsoft 365 Copilot, respect defined data boundaries.
• Data Governance and Compliance Tools include solutions for data mapping, access control, audit logging (like Data Guardians tamper-evident ledger), and real-time monitoring of data residency.
• Security Information and Event Management (SIEM) is important for comprehensive security monitoring and incident response within sovereign environments.
• Partner Integration Frameworks are platforms that facilitate seamless collaboration with local cloud partners and service providers.

Key Performance Indicators (KPIs)

Key Performance Indicators (KPIs) focus on metrics that reflect genuine public engagement and AI governance.

• Track Public Adoption Rate, measured as the percentage of citizens and public institutions actively using CanGPT services.
• Monitor Democratic Engagement Score, which covers metrics on public participation in defining AI governance, content moderation, and ethical guidelines.
• Assess Energy Efficiency Ratio, the computational output per unit of energy consumed, particularly for targeted tasks.
• Measure Data Diversity and Representation Index, which tracks the inclusivity and representativeness of the training data used by CanGPT.
• Finally, evaluate Public Value Creation through qualitative and quantitative assessments of how CanGPT contributes to social good, civic engagement, and access to information.

A structured Review Cadence

A structured Review Cadence is also critical.

• Conduct Quarterly Public Consultations, regular forums for citizen input on AI features, policies, and ethical concerns.
• Engage in Bi-Annual Technical Performance Reviews to assess CanGPTs performance, update mechanisms, and resource utilization.
• Finally, perform an Annual Independent Ethical Audit to conduct external audits of AI governance, bias detection, and compliance with democratic principles, and establish Ongoing Open-Source Contribution Monitoring to track and encourage community contributions to the CanGPT platform.

FAQ

Question: What is CanGPT?

Answer: CanGPT is a proposed national, public-service AI model for Canada, inspired by public broadcasters like the CBC, aiming to serve public values rather than solely commercial interests.

Question: Why is a public AI model being considered in Canada?

Answer: It is being considered to ensure AI serves the public amid calls for a public interest approach to AI policy, to democratically define AIs ethical boundaries, and to offer a contrast to commercial AIs environmental impact and reliance on public data without contributing back.

Question: How would CanGPT be different from commercial AI like ChatGPT?

Answer: CanGPT would be built as a public utility, potentially open-source, drawing on public domain materials and government datasets.

It would aim for democratic governance over content moderation and prioritize energy efficiency, contrasting the private, commercially driven models.

Glossary

• CanGPT: A conceptual national, public-service Artificial Intelligence model for Canada, envisioned as a public utility.
• Public AI: Artificial Intelligence systems developed and governed by public institutions, prioritizing public good over commercial gain.
• Digital Sovereignty: A nations ability to control its data, digital infrastructure, and online activities within its own borders.
• Generative AI: A type of Artificial Intelligence that can create new content, such as text, images, or audio.
• AI Governance: The framework of rules, policies, and processes for guiding the design, development, and deployment of AI systems.
• Content Moderation: The process of monitoring and filtering user-generated content to ensure it complies with established rules and guidelines.
• Knowledge Commons: Shared information and knowledge resources that are collectively owned or managed, like the internet.
• Public Utility: A service or resource (like electricity, water, or a public broadcaster) essential for public welfare, often regulated or provided by the government.

Conclusion

The conversation around AI is too vital to be left solely to market forces.

For Canada, the idea of CanGPT is more than a novel technical project; it is a profound ethical reflection.

It asks us to look beyond the slick interfaces and impressive capabilities of commercial AI and consider what kind of digital future we truly want to build—one shaped by the logic of profit, or one guided by the enduring values of public service, democratic participation, and collective well-being.

By fostering a national conversation and committing to a public AI model, Canada has the opportunity to redefine what digital sovereignty and AI innovation can mean, proving that the most powerful technologies can indeed be built for all.

The answer to ChatGPT might not be another subscription, but a renewed commitment to the public good.

References

Could a national, public CanGPT be Canada’s answer to ChatGPT.