Skip to main content
Bureau of AI
Category: Trustworthy AI Principles

Sustainability

Simply put

Sustainability is the ability to meet present needs while preserving the capacity to meet future needs over the long term. It is commonly described as balancing environmental health, social equity, and economic vitality so that the resources and conditions people depend on are not depleted. As commonly used, the term spans near-term operations and long-term stewardship rather than referring to any single practice.

Formal definition

Sustainability, as commonly defined across the evidence, refers to the capacity to maintain or improve desirable materials, conditions, and processes over time without depleting natural resources. It is frequently framed as the integration of three interdependent dimensions: environmental health, social equity, and economic vitality. Definitions vary by source and institution and are not standardized; the term should be scoped to the domain in which it is applied, and the evidence provided does not address AI-specific or model risk management usage.

Why it matters

Sustainability has become a central organizing concept across public policy, corporate strategy, and institutional planning because it frames decisions in terms of long-term consequences rather than near-term gains alone. As commonly described in the evidence, it rests on a simple principle: much of what people need for survival and well-being depends, directly or indirectly, on natural resources and conditions that can be depleted. This makes sustainability a lens for evaluating whether current activities can continue without undermining the capacity of future generations to meet their own needs.

The concept matters in part because it is frequently framed as the integration of three interdependent dimensions—environmental health, social equity, and economic vitality—rather than as an environmental concern alone. This multidimensional framing means that trade-offs between economic performance, social outcomes, and ecological limits are made explicit, which is why sustainability language appears in strategy, reporting, and stewardship contexts.

A practical caution is that definitions vary by source and institution and are not standardized. The term should be scoped to the domain in which it is applied, and readers should be aware that the evidence provided here addresses sustainability as a general concept and does not cover any AI-specific or model risk management usage of the term.

Who it's relevant to

Policy and public-sector specialists
Those working in environmental and public policy encounter sustainability as a principle grounded in the dependence of human survival and well-being on natural resources. It provides a framework for evaluating whether current activities can continue over the long term without depleting those resources.
Corporate strategy and reporting professionals
For those involved in organizational strategy and disclosure, sustainability is commonly framed as maintaining or supporting economic, environmental, or social processes over time without depleting natural resources. Because definitions are not standardized, professionals should scope the term to the domain and institution in which it is being used.
Institutional and community planners
Those planning for communities and organizations use sustainability to consider needs not only in the near term but over the long term. The concept's framing around environmental health, social equity, and economic vitality supports planning that weighs these dimensions together.

Inside Sustainability

Environmental sustainability of AI systems
The consideration of energy consumption, computational resource use, and associated carbon footprint arising from training, deploying, and operating models. In many frameworks this is treated as one dimension of responsible AI rather than a standalone regulatory requirement.
Operational sustainability (lifecycle continuity)
The organizational capacity to maintain, monitor, retrain, and support a model over its full lifecycle. This overlaps with model risk management concerns such as ongoing monitoring and performance degradation, but is distinct from environmental sustainability and should not be conflated with it.
Governance sustainability
The durability of the policies, roles, accountability structures, and oversight processes that support AI systems over time. This is an AI governance concern (organizational structures and oversight) rather than a measurement-of-model-risk activity, though the two overlap where governance sustains monitoring controls.
Resource and cost sustainability
The economic viability of sustaining data pipelines, infrastructure, staffing, and validation activities needed to keep a model in reliable operation. Typically framed as an operational consideration rather than a binding regulatory obligation.

Common questions

Answers to the questions practitioners most commonly ask about Sustainability.

Does 'sustainability' in an AI governance context mean the same thing as environmental sustainability?
Not necessarily. In AI governance discussions the term is used in more than one sense, and professionals frequently conflate them. It can refer to the environmental footprint of AI systems (for example energy and compute consumption associated with training and inference), but it is also used to describe the operational sustainability of a model or governance program, meaning its ability to remain fit for purpose, adequately resourced, and controlled over time. Because the term is contested and context-dependent, you should specify which meaning you intend rather than assume a single definition applies across all frameworks.
Is model sustainability just another name for model performance?
No, and treating them as interchangeable is a common error. Performance typically refers to how well a model meets its accuracy or effectiveness metrics at a point in time, whereas sustainability, as commonly used, concerns whether a model can continue to be maintained, monitored, resourced, and controlled appropriately over its lifecycle. A model can perform well at deployment yet be unsustainable if there is no ongoing ownership, monitoring, or capacity to manage its risks. These concerns overlap but should not be collapsed into one.
How can an organization operationalize sustainability considerations within an existing model risk management program?
Organizations commonly address sustainability by embedding lifecycle considerations into existing controls rather than creating a separate structure. This can include assigning clear ownership for ongoing maintenance, defining monitoring and review cadences, documenting resourcing needs, and establishing criteria for when a model should be revalidated, redeveloped, or retired. Whether these are formal requirements depends on the applicable framework and sector, so scope them to your governance policies and any binding obligations that apply to you.
Where do responsibilities for sustainability typically sit across the lines of defense?
In many organizations, responsibilities are distributed rather than held by a single function. The first line (model owners and developers) is typically responsible for building and maintaining models in a sustainable manner, the second line (independent risk or validation functions) generally challenges and reviews whether sustainability considerations are adequately managed, and the third line (internal audit) may assess the effectiveness of those controls. The specific allocation should follow your organization's defined operating model, since roles vary by institution and sector.
What documentation supports demonstrating that sustainability considerations are being managed?
Documentation commonly used includes model inventories, ownership and accountability records, monitoring plans and results, review or revalidation schedules, and records of decisions to retire or replace models. The appropriate documentation depends on the framework and expectations that apply to your context, so align it with your governance policies rather than assuming a universal standard set.
How does sustainability relate to model retirement and decommissioning decisions?
Sustainability considerations often inform when and how a model is retired, because a model that can no longer be adequately maintained, monitored, or resourced may need to be decommissioned or replaced. Retirement is generally treated as part of managing model risk across the lifecycle rather than as a separate discipline. The specific triggers and procedures should be defined in your governance and model risk policies, as practices vary across organizations and sectors.

Common misconceptions

Sustainability in AI refers only to environmental or carbon-related impacts.
As the term is commonly used, sustainability can refer to environmental impact, operational lifecycle continuity, or the durability of governance structures. These are distinct dimensions and should not be collapsed into a single environmental meaning.
Sustainability requirements for AI are settled, universally binding regulatory obligations.
Regulatory treatment of AI sustainability is evolving and varies by jurisdiction and instrument. Frameworks differ in whether they are binding law, guidance, or voluntary standards, and specific sustainability requirements should not be presented as universally applicable settled law.
Sustaining a model operationally is the same as managing its model risk.
Operational sustainability (keeping a model running, retrained, and supported) overlaps with model risk management activities such as ongoing monitoring, but the two are distinct. Sustainability addresses continuity and resource capacity, while model risk management focuses on identifying, measuring, monitoring, and controlling risks arising from model use.

Best practices

Clarify which dimension of sustainability is intended (environmental, operational, or governance) in any policy or documentation to avoid conflating distinct concerns.
Where sustainability requirements are cited, identify the issuing body and whether the instrument is binding law, guidance, or a voluntary standard, using qualified language where certainty is lacking.
Integrate operational sustainability considerations with ongoing monitoring processes so that lifecycle continuity supports, but does not replace, model risk management controls.
Assign clear accountability for sustaining governance structures over time, recognizing this as an AI governance responsibility distinct from measuring model risk.
Document the resource, cost, and staffing assumptions required to sustain a model, and revisit them as conditions change, framing these as risk-reducing measures rather than guarantees.
State the scope and limitations of sustainability commitments explicitly, noting where definitions are contested or where treatment differs by sector or jurisdiction.