The GenerIA Blog
Regulating Frugal AI: Between Progress and Challenges...
Frugality is a radical shift in the way businesses and governments think about AI. But how do we regulate a technology that promises both performance and a sustainable environmental footprint? Let's take a look at how three major regions - Canada, Europe and the United States - are approaching the problem...
In Canada, the prevailing consensus is that mankind has only a limited supply of natural resources. It's hardly surprising, then, that our country sees frugal AI as essential to reconciling technological innovation and eco-responsibility. Although the legislative framework for AI does not yet include specific regulations on frugality, concrete measures are being taken to manage the impact of new technologies and encourage sustainable practices.
One example is the Standards Council of Canada's (SCC) Verification and Validation Accreditation Program, which aims to guarantee the accuracy and reliability of companies' environmental declarations. It verifies the sustainability efforts of technologies and ensures that ecological approaches are not greenwashing. In other words, it validates that companies comply with strict standards in terms of energy consumption and resource management, particularly in the field of AI. This program is vital for differentiating genuinely green actions from purely marketing-oriented claims. As such, it is destined to play a crucial role in the adoption, hopefully in the near future, of environmental standards for artificial intelligence.
At the same time, Bill C-27, which aims to regulate AI in Canada, is beginning to include ethical principles concerning data protection and the transparency of algorithms. While the legislation does not yet directly address frugal AI, it does set out a broader legislative framework for regulating emerging technologies. It is likely that this draft, once adopted, will evolve to include environmental criteria, particularly with regard to the energy consumption of technologies and the ecological impact of AI.
These two initiatives demonstrate Canada's commitment to responsible technology, where sustainability and transparency are not left to chance or to the goodwill of suppliers and customers. As frugal AI continues to develop (thank you GenerIA!), these validation and regulatory mechanisms could provide a model for other countries, particularly in terms of how they might integrate environmental criteria into their technology policies.
Europe in the vanguard of standards
The European Union has established itself as a reference for AI regulation, and frugal AI is no exception. In 2025, Europe is continuing to expand its efforts to regulate the most resource-intensive technologies. The AI Regulation, which provides a framework for the use of AI at European level, places particular emphasis on the environmental impact of the technologies deployed, and is intended become a formal incentive. While frugal AI has not yet been specifically named, this legislative framework paves the way for real sustainability standards.
For the EU is considering imposing stricter requirements on the energy consumption and carbon footprint of systems. These requirements should be based on criteria relating to the reduction of CO2 emissions and water stress, the management of equipment obsolescence and the reuse of resources.
The French initiative: AFNOR and the Reference Framework on Frugal AI
In France in particular, the most significant initiative in terms of frugal AI is the AFNOR Reference Framework on Frugal AI which provides a proactive and ambitious framework for companies wishing to adopt eco-responsible digital solutions. The standard offers a structured approach to integrating sustainability into the design, deployment and use of AI technologies. As well as encouraging the optimisation of material resources and energy consumption, the reduction of electronic waste and the carbon footprint of the infrastructures used, it promotes training, awareness-raising and support programs (depending on the target audience), without which, quite reasonably, the game will be hard to win.
At the same time, a television advertising campaign run by ADEME (the French Agency for ecological transition) in partnership with AFNUM (the French Agency for everything digital, based on the AltIMPACT Program, is evangelizing a more sustainable vision of digital technology and specifically proposing the adoption of frugal AI practices. Initiatives such as these are essential to raise awareness among the population (from decision-makers in companies to end consumers on their mobile phones) of the importance of reducing energy costs and preferring environmentally-friendly technologies.
It is conceivable that the AFNOR initiative could serve as a model for the introduction of specific European texts on frugal AI, particularly as regards reducing emissions and optimizing resources in low-cost systems. In addition to the legislative framework of the European AI Act, increased collaboration between EU countries, AFNOR and its national partners, and the ecosystem could lead to the introduction of common standards for frugal AI. For their promoters, these regulations will offer ethical and sustainable solutions while meeting the growing demand for more accessible technologies.
The US paradox: innovation vs. sustainability
In the United States, the environmental regulation of AI, which was still embryonic under the Biden administration, could take a real step backwards under the Trump administration. Private initiatives are certainly beginning to emerge, particularly in sectors where profitability and energy efficiency are economic priorities. But these efforts remain discretionary, and are not accompanied by any regulatory framework or incentives for ecological responsibility.
Given the importance of AI to the country, and the country's importance in the global AI ecosystem (both in research and production), this lack of framework poses a problem. At current rates, whatever the metrics, a rapid attrition of resources is in view. Already, in the States with the largest number of AI data centers, there has been a real deterioration in basic services (as in the technical quality of electricity, for example, which can lead to irreversible damage to the devices themselves), but what's more, the cost of these services is rising to the point of favouring businesses over households.
The major players all claim to be sourcing 'green energy', while at the same time developing exclusive internal sourcing programs that are sometimes based on predatory practices. The fact remains that the equation will have to be resolved at national level: without a binding shift towards frugality, American AI faces an existential threat, from innovation to day-to-day production.
Conclusion
To be sustainable, AI must be frugal. Regulating it, in order to encourage it, is more a matter of necessity than virtue. Can we therefore envisage the emergence of proactive international agreements, with institutions such as the G7 or the OECD unanimously deciding to apply common standards in a concrete and practical way? Given the current geopolitical landscape, it is difficult to be optimistic. The reasons for hope are more likely to come from China, where the most recent general-purpose AI systems demonstrate that, even when severely constrained, frugality does not prevent excellent performance...
In the GenerIA blog:
