By Kevin Frazier
Kevin Frazier is an Assistant Professor at St. Thomas University College of Law. He also serves as a Director of the Center for Law & AI Risk. Professor Frazier appreciates the feedback he received on this paper from AI Futures Fellowship participants. Finally, he is grateful to the Journal of Law and Technology at Texas staff for their work on this article and the assistance of his Research Assistant, Ana Barreto.
Summary
(footnotes omitted)
The Forest Stewardship Council (FSC) deserves study by AI safety advocates. The FSC came about as a result of states being unable to rally behind a “hard law” solution to management of the world’s forests. Though the FSC lacks formal legal authority, it has nevertheless contributed to creating and enforcing more sustainable forestry practices by soliciting a high degree of buy-in from a diverse range of actors. As the leading certifier of sustainable forestry practices, the FSC’s innovative institutional design is particularly worthy of analysis. Each governing body within the FSC, such as the organization’s General Assembly and assemblies at the regional and national levels, affords equal voting power to three chambers—one each for representatives from economic, environmental, and social organizations. Within each chamber at the international level, sub-chambers of members from the “North” and “South”—defined by economic size rather than geographical location—wield equal voting power.
These structural safeguards do not always work in practice. Members from the economic sector and from the North tend to have more resources to allocate toward shaping and monitoring FSC standards and processes. The upshot is that even innovative and inclusive governance structures will not lead to the intended substantive outcomes if thought is not given to resource disparities and differences in the strengths of incentives among members.
The FSC relies on market-based incentives to further its regulatory aims. This reliance has severe limitations. The first limitation arises from the paradox of trying to stem deforestation by encouraging the sale of “good” timber. That is, in the same way that those invested in protecting Alaska’s salmon might feel uneasy about promoting more consumption of “sustainably-caught” fish, the FSC’s scheme can place individuals aiming to preserve forests in the odd position of increasing demand for timber. This dynamic has troubled environmental and social members who may favor greater limitations on the sale of any sort of timber. The second limitation arises from competition with other certification efforts. More “timber-friendly” certification efforts have facilitated a race to the bottom with respect to the development and enforcement of strict standards; sadly, the FSC has joined this race.
AI safety advocates weighing whether to develop a “safety” certification for AI models should take away a few key lessons from the FSC. The first lesson is that institutional design matters. Though the FSC’s novel and inclusive structure rightfully received praise upon the organization’s creation, the subsequent realization of shortcomings associated with that structure should have led to responsive institutional reforms. Yet, such reforms have gone unexplored or unrealized. As a result, the power parity that was supposed to exist among all global and sectoral actors has been undermined. Instead, stakeholders with more resources and vested interests have seized control of the FSC.
The second lesson is that market-based mechanisms may not be the right approach if stakeholders want to decrease the availability of even “good” or “less bad” products. Relatedly, market-based mechanisms increase the odds of a “race-to-the-bottom” dynamic that results in profit-motivated stakeholders having more leverage over the creation and enforcement of standards.
Finally, the third lesson is that consumer awareness of and education related to “good” products is essential to the success of a certification system. Consumers have played a leading role in holding the FSC and its members accountable.