When Environmental Governance Fails, What Is Left for AI?

By Javier Surasky



In March 2022, the United Nations Environment Assembly, meeting in Nairobi, adopted the resolution End plastic pollution: towards an international legally binding instrument, initiating a process aimed at adopting a legally binding treaty against plastic pollution. The negotiations unfolded over six rounds, beginning in Uruguay and continuing in France, Kenya, Canada, South Korea, and, until yesterday, Geneva.

But what many had already foreseen eventually happened: the negotiations collapsed yesterday when representatives from 184 countries gathered in Geneva failed to agree on issues such as establishing mandatory limits on plastic production and on the plastic industry’s use of toxic chemicals.

The initial feeling is one of frustration, but after that first moment, questions arise that go beyond the failed agreement itself. I wonder, for instance, what opportunities this regulatory vacuum presents in terms of using digital technologies in the fight against climate change.

Although this question may seem secondary compared to the significance of the negotiations’ failure, no one can now doubt that AI can, and indeed already does, be an ally in combating plastic pollution: from using neural networks to identify plastic waste in marine ecosystems through satellite imagery, to predictive models on plastic accumulation and its impacts, these tools have great potential to guide decisions and actions in the fight against plastic pollution.

Let us take some recent examples: deep learning algorithms have not only managed to classify plastic waste with an accuracy of 90 to 98%, but have done so more efficiently than manual sorting systems, while ConvoWaste combines computer vision, sensors, and actuators to separate waste with 98% accuracy. Other systems have been designed to sort common types of plastic in portable devices, achieving similar levels of effectiveness. This demonstrates how AI can contribute to automated recycling processes and the design of more efficient public policies for managing plastic pollution and waste.

For these tools to reach their full potential, they require institutional frameworks that embed them within coordinated strategies involving multiple stakeholders. This is where the treaty’s failure strikes hard: without international agreements that promote the use of AI for the common good, and without arrangements to work jointly against microplastics, the two pillars needed to scale up these advances collapse.

The disconnect between environmental governance and digital governance exacerbates the problem. There is a lack of intersections and cross-fertilization between debates on how to limit plastic production and on regulating AI. This results in inefficient and dangerous fragmentation.

We already know that AI is not a neutral technology. It can be implemented following extractivist logics, maximizing economic gain while externalizing environmental costs. Still, it can also be deployed with a focus on sustainability and the promotion of human rights, especially those of the most vulnerable, if integrated with ethical frameworks oriented toward environmental justice.

The difference lies, first, in the assignment of purposes to AI, but equally important is the consideration of the data that feed its models, something that becomes central in the specific case of plastic pollution.

Collecting data on plastic waste, detecting microplastics in different ecosystems, modeling their circulation, or understanding their impacts on human health are all areas where intelligent systems are increasingly present. However, so far, these systems are mainly developed by private actors, with varying levels of transparency and under methodologies that do not include guarantees of democratic access to training data or to the data generated through processing.

Who decides today what data is collected? Who has access to them and under what conditions? Who validates them and under what pressures? These are questions that the development of AI for Sustainable Development must address, and they directly affect the field of combating plastic pollution. Data related to production, distribution, composition, and toxicity of plastics are fragmented, of limited accessibility, or outright inaccessible. Most of them are in the hands of companies under no binding obligation to share them. Moreover, these companies use AI as a tool to craft their corporate lobbying strategies: analyzing political trends, pushing disinformation campaigns, segmenting public opinion, and influencing national decision-makers who then bring these positions to multilateral debates. AI is not just a valuable tool for solving environmental problems; in the hands of those who benefit from the current situation, it serves to preserve the status quo.

As a result, we have sufficient technological advances to address the microplastics problem, but we lack international cooperation frameworks that foster their deployment under equitable conditions.

It is no coincidence that the failure to agree on a text in Geneva was due, in significant, though hard to quantify, measure, to opposition from certain countries and corporate lobbying by those benefiting from mass plastic production and the use of chemical additives. In short, the debates revealed two major coalitions of states:

  • A large majority bloc (the EU, Pacific island states, most governments in Asia and Latin America and the Caribbean) advocated for an agreement that would set global, mandatory limits on plastic production, including the elimination of single-use plastics and the prohibition of potentially dangerous chemicals. This group was supported by global civil society.
  • A second bloc, comprising major oil-producing powers and the petrochemical industry (with over 220 petrochemical company lobbyists registered to participate in the meeting), was led by China, Saudi Arabia, the United States, and Russia. This bloc sought a treaty focused on improving waste management and promoting recycling.

In many cases, these states and companies lead the development of AI-based solutions, which raises the question: Can algorithmic tools help solve a problem if they are developed and controlled by those with vested interests in its persistence?

The answer, intuitively negative, brings us back to the link between the environmental and digital spheres. There can be no environmental justice without algorithmic justice, and neither is achievable without a new, more inclusive and democratic model of multilateral governance. Viewed this way, the intersection between environmental governance and algorithmic governance is a prerequisite for addressing the challenges posed by climate change.

For all these reasons, we see the failure of the plastics treaty as a wake-up call about the urgent need to build a public, global digital framework aimed at solving shared problems.

After all, the production of plastics and plastic waste, as well as the major challenges for developing AI for good, share several root causes: extractivism, opacity, insufficient regulation, and the lack of long-term vision among decision-makers.

If we believe AI can help solve the planet’s most significant problems, the time has come to ensure it is part of the conversation from the start, and to build the legal frameworks and consensus needed for its deployment.