From Taboo to Assumed: How Carbon Dioxide Removal Advanced Before Governance Could Keep Up and What It Means for SRM

This blog post is one in a series drawing out lessons for solar radiation modification from efforts to govern other emerging technologies. The goal is to seek insight into how institutions and societies can best prepare for the complex task of managing powerful and disruptive technological innovations. The introduction to the blog series is here.

Carbon dioxide removal (CDR) was not always regarded as a respectable policy option. For many years, CDR proposals surfaced many of the same concerns associated with climate adaptation in its early days: it was seen as accepting defeat on mitigation goals, providing political cover for continued emissions, and solving the wrong problem. The mitigation deterrence argument shaped what researchers could credibly propose and what policymakers could openly consider. Addressing legacy emissions and the likelihood of temperature overshoot would require not just stopping emissions, but actively removing carbon already in the atmosphere. CDR shifted from an unwelcome idea to a necessary tool, not because the political objections disappeared, but because staying within safe limits left fewer and fewer alternatives. By the time political bodies were formally engaging with CDR, its role in 1.5°C pathways had already been established through modeling choices, and the governance conversation had to play catch-up to assumptions that had quietly hardened into baselines. That shift from taboo to assumed matters because it reveals how easily technical and political momentum can outpace the institutions meant to guide it.

The debate about whether CDR and solar radiation modification (SRM) should be understood as fundamentally different forms of climate intervention, or placed under the same geoengineering umbrella, carries real governance implications. There are some very key differences across these approaches. Fundamentally, CDR poses a collective action problem: it requires many actors, substantial capital, and coordination across decades to deliver climate benefits that no single actor can capture on its own. SRM raises a different set of questions, and in some ways, harder ones. The core governance challenge is not just the risk of unilateral action, but what unilateral action would mean in practice: a small number of actors making decisions about the global climate system that could affect rainfall patterns, agricultural yields, and temperatures in places far removed from where deployment decisions were made. Who gets to set the global thermostat, on whose terms, and with what accountability to those who bear the consequences, are questions that no existing institution is equipped to answer. 

But both technologies also have important governance commonalities. They both create the need for long-term considerations and policies: CDR requires sustained effort to deliver cumulative removals and durable storage, while some forms of SRM— particularly stratospheric aerosol injection at scale — cannot easily be quickly stopped without risking rapid and potentially catastrophic rebound warming, a phenomenon known as termination shock. Both approaches have potential benefits, but both also involve interventions under uncertainty, and carry risks of mitigation deterrence. The governance of a slow-moving, capital-intensive climate intervention like CDR tells us something useful about what happens when governance lags — particularly for a fast, relatively cheap intervention where the absence of adequate frameworks risks leaving consequential decisions to a small group of powerful actors moving ahead without broad agreement.

Proceeding in a vacuum: the example of ocean fertilization 

The clearest illustration of what research governance failure looks like in practice comes from ocean iron fertilization, an early and instructive case of CDR research proceeding in a regulatory vacuum. The basic idea of seeding parts of the ocean with iron could trigger phytoplankton blooms capable of absorbing atmospheric carbon attracted scientific interest in the 1990s and early 2000s. Small-scale experiments were conducted, mostly by researchers operating in good faith, but in the absence of any agreed-upon framework for what oversight should look like, who had authority to authorize experiments, or how affected communities should be consulted. The governance gap became impossible to ignore in 2012, when a private entrepreneur named Russ George dumped approximately 100 tonnes of iron sulfate into the Pacific Ocean off the coast of British Columbia, framing it as both a carbon offset scheme and a salmon restoration project. The experiment had not been subject to meaningful scientific review, had no monitoring framework capable of assessing its impacts, and consent was narrow, procedurally questionable, and not representative of the range of communities with legitimate interests in the outcome. The backlash was swift, and the incident became a defining moment in how the CDR community understood the relationship between scientific legitimacy and public trust. Months after this incident, the London Protocol parties adopted a resolution establishing a regulatory framework that restricts ocean fertilization to legitimate scientific research, a reactive constraint that arrived after the damage to public trust had already been done. 

What the ocean iron fertilization case reveals is not simply that bad actors can exploit governance gaps, though they can. It reveals something more structural: that when oversight frameworks are absent, even well-intentioned research struggles to build the legitimacy it needs to continue. The scientific community lost ground not because the underlying science was discredited, but because the process through which experiments were conducted failed to earn public trust. Similar dynamics are beginning to shape the SRM space: technical actors moving ahead of governance actors, triggering controversy that narrows what is politically possible and where techno-optimism has sometimes obscured the urgency of building oversight frameworks before capability outpaces accountability.

Technology ahead of governance

That vacuum didn't form overnight. CDR's rise in mainstream climate policy was driven by genuine necessity: as emissions reduction targets became more ambitious, growing consensus consistently concluded that limiting warming to 1.5°C would require removing carbon from the atmosphere at scale, in addition to robust mitigation. But governance arrived late and investments started pouring in. The US Inflation Reduction Act (IRA) and Infrastructure Investment and Jobs Act (IIJA) channeled unprecedented public funding into CDR scaling, accelerating timelines before international governance infrastructure had been built to match. Private actors moved in the same direction. Where binding frameworks were absent, voluntary mechanisms emerged to fill the gap, with consequences that the CDR community is still feeling today.

These cases share a common pattern. Governance frameworks did not fail because no one cared about oversight; they failed because the pace of technical development and investment consistently outran the institutions meant to guide them. By the time robust frameworks were being seriously discussed, activity had already taken shape, actors had already staked out positions, and assumptions had already hardened into baselines that were difficult to challenge. CDR's trajectory illustrates how some technologies advance not through deliberate decision-making but through quiet momentum: each technical step small enough to seem manageable, the cumulative effect large enough to foreclose options that were never explicitly debated. 

Voluntary governance mechanisms can fill that gap, but their design and ownership matter as much as their existence. Frameworks created by and for commercial actors, where the entities responsible for verification are paid by those whose project they assess and where revenue depends on the volume of certified activities, face structural incentive pressures that frameworks created by scientific, civil society or governance actors do not. When commercial pressures dominate, voluntary frameworks tend to generate the appearance of oversight rather than produce it. That is a structural problem, not an incidental one.

The pattern across these cases points to a consistent sequencing problem: technical and commercial actors move first, investment follows, and governance actors are left catching up to decisions that have already shaped the terrain. In CDR's case, that sequencing was costly but the damage remained bounded; the window for course correction, however imperfectly used, stayed open.

SRM compresses that window dramatically. SRM can be developed at dramatically shorter timescales, with less infrastructure requirements, and is inherently globally scaled in impacts. The realities of climate vulnerability mean that the costs of getting governance wrong would likely not be distributed among those who made the decisions. And the low cost of entry means that the number of actors capable of moving ahead without broad agreement is growing, not shrinking.

The CDR analogy is useful not because the technologies are equivalent, but because it shows what happens when the sequence goes wrong. Technical momentum precedes deliberation, technological development outpaces the governance actors who should be shaping it. For CDR, that sequence produced fragile markets, damaged trust, and frameworks built around damage limitation. For SRM, allowing the same sequence to repeat would mean learning the wrong lesson from history. Effective SRM governance will need to ensure that collective frameworks are in place before technical and political momentum narrows the options, not after the vacuum has already been filled by actors with the means and incentives to move first.

CDR is one case in a longer pattern. The piece that follows examines how similar dynamics played out with AI illuminating a different dimension of what governance readiness requires, and what it has cost when it arrives too late.