Geoengineering Governance

Governance here means the rules determining who may research, test or deploy a deliberate climate intervention, under what conditions and with what accountability. It matters disproportionately because of two properties established in Climate Engineering: solar radiation modification is fast and it is cheap. A technology requiring a trillion dollars and thirty years governs itself through its own difficulty. One a single state could attempt inside a decade, at roughly $2.25 billion a year, does not.

This brief takes the institutional questions and none of the physics. Deployment costs, model results, termination shock and the outdoor experiment record belong to the climate engineering brief; the ozone regime's own history and compliance machinery belong to Atmospheric Management; the general question of whether any institution can manage a planet belongs to Planetary Stewardship. What is left is a specific and unusual pathology: an intervention that costs less than the institution needed to govern it. That does not appear in the stewardship framing and it is this slot's spine.

The framing under test is that the governance can be built before the deployment. Established The record contains a clean natural experiment on exactly that proposition and it came out the wrong way. Between 2019 and 2026 the international system tried twice to establish the mildest possible instrument and failed both times; the one adopted amendment addressed to a climate-intervention technique has nine acceptances after thirteen years and is not in force. In the same window a startup conducted more than 124 stratospheric sulfur releases, another company raised $75 million on a deployment slide, and two US states enacted the world's only binding prohibitions — largely on the strength of a belief that is false.

Established There is no binding international instrument governing solar radiation modification research, outdoor experimentation or deployment. That is the starting fact and it has not changed. What exists instead is a set of instruments that bind partially, bind adjacent conduct, or do not bind at all — and the discourse routinely treats the weakest of them as the strongest. Setting them out in order of actual force is the most useful thing this brief can do.

Established ENMOD is the most-cited and least-applicable instrument in the field. The Environmental Modification Convention entered into force on 5 October 1978 and has 78 states parties. Article I prohibits military or any other hostile use of environmental modification techniques having widespread, long-lasting or severe effects; Article II defines the technique class expansively, as any technique for changing, through deliberate manipulation of natural processes, the dynamics, composition or structure of the earth, including its biota, lithosphere, hydrosphere and atmosphere, or of outer space. Established Three things gut it here. It applies only to hostile use, and a state deploying aerosols to cool its own climate is not conducting hostile environmental modification however severe the transboundary effect. The troika threshold — widespread, long-lasting or severe — sets a high bar, and attempts to remove those qualifiers failed for want of consensus at the original negotiations and at subsequent review conferences. And 78 parties is a minority of states. ENMOD is frequently invoked as the geoengineering treaty and it governs almost none of the conduct at issue.

Established The London Convention and London Protocol are the one place a specific climate-intervention technique has genuinely been regulated, and almost nobody reports what is happening there now. Resolution LC-LP.1 (2008) established that ocean fertilisation other than legitimate scientific research should not be allowed; LC-LP.2 (2010) added an assessment framework for judging what counts as legitimate research; and Resolution LP.4(8), adopted 18 October 2013, added Article 6bis, an Annex 4 listing marine geoengineering activities — ocean fertilisation only — an Annex 5 assessment framework, and a treaty definition of marine geoengineering as a deliberate intervention in the marine environment to manipulate natural processes, including to counteract anthropogenic climate change, with potential for deleterious effects. Established It is not in force. As of 2025, nine parties have accepted it: Australia, Croatia, Denmark, Estonia, Finland, Germany, the Netherlands, Norway and the United Kingdom. Thirteen years, nine acceptances. The 2008 and 2010 resolutions continue to apply pending entry into force, but those are resolutions rather than treaty obligations.

Established And the live process there is the most consequential fact in this section. At the 2025 meetings, parties expressed concern about the growing number of marine geoengineering activities by private and commercial entities and their potential impact on developing nations, and established an intersessional correspondence group running to October 2026 to clarify how existing instruments apply, refine definitions for priority techniques, and support implementation of assessment frameworks. Ocean alkalinity enhancement, biomass sinking, artificial upwelling and marine cloud brightening are among the techniques under consideration for listing. Frontier The London Protocol is therefore the one existing treaty body with a live process that could bring a solar radiation modification technique under binding international regulation. Not the UNFCCC, not the UN Environment Assembly, not the Convention on Biological Diversity. A marine dumping convention.

Frontier The CBD moratorium is not a moratorium. Decision X/33 (2010) invited parties to ensure that no climate-related geoengineering activities affecting biodiversity take place absent adequate scientific basis and consideration of risks, with an exception for small-scale controlled scientific research studies; COP16 in 2024 reaffirmed the precautionary approach. It is universally described as a de facto moratorium, and what it actually is, is a COP decision under a framework convention, addressed to parties as an invitation, with a research carve-out, and with the United States not a party. Its normative weight is real and its legal force over a determined actor is close to nil, and the gap between those two things is where most of the confusion in this subject lives.

Frontier The ICJ's advisory opinion of 23 July 2025 is the newest and most general instrument, and its significance is not what most commentary says. It is an advisory opinion, so it binds nobody directly. Its importance is that it articulates obligations — prevention of significant transboundary harm, due diligence, regulation of private actors — as existing law rather than new law, which means they applied to climate intervention before anyone drafted for it. No court has tested that against an SRM actor, and until one does the proposition is a reading rather than a holding.

Established Nationally, the position is oblique everywhere and explicit nowhere. The US EPA's own page, last updated April 2026, states plainly that Congress has not passed any law solely related to solar geoengineering. What exists instead is the Weather Modification Reporting Act of 1972, requiring notification to NOAA at least ten days before activities including release of gases, dusts, liquids or aerosols into the atmosphere; the Marine Protection, Research and Sanctuaries Act, which may require permits for ocean-related activity and did for the 2025 ocean alkalinity trial; Clean Water Act section 402 for some marine discharges; and the Clean Air Act, under which the EPA used section 114 to demand information from a balloon-launching startup. NOAA is the only federal agency with explicit SRM research funding since 2020.

Established The 2024 UN Environment Assembly attempt is the closest the international system has come, and the way it collapsed maps the political geography better than any commentary. Switzerland led a resolution to establish a UNEP expert group assessing the science and possible governance of SRM. The United States, supported by Saudi Arabia and Japan, resisted the inclusion of governance, social, legal, geopolitical and security risk assessment, preferring a purely scientific evaluation routed through the World Meteorological Organization. The African Group, represented by Djibouti, went the other way: invoking the August 2023 African Ministerial Conference on the Environment call for a global governance mechanism for the non-use of SRM, they sought moratorium language, broader knowledge-sharing, risk-centred framing and reference to the 2010 CBD decision. Brazil, Mexico, Colombia, Pakistan, Fiji, Vanuatu and Barbados largely aligned with them; the EU sought broader assessment and precautionary language; Norway, the UK, Australia and Canada took softer positions. Switzerland withdrew the resolution after successive compromises had reduced it to a proposal that UNEP prepare a paper on establishing an information repository — and even that proved divisive.

Established Three findings follow and all three cut against the framing. First, the two poles were not “for” and “against” SRM; they were “science only, no governance” and “governance first, specifically non-use.” The states most able to deploy resisted governance; the states most likely to be affected demanded it — the exact inverse of the arrangement that would let governance precede deployment. Second, the failure point was not deployment authorisation but an information repository: a body that would have done nothing but collect what is already public could not be agreed. Whatever the international system's capacity for SRM governance is, it is below the threshold of a filing cabinet. Third, this was the second failure, not the first — a similar Swiss-led attempt collapsed at UNEA-4 in 2019. Two attempts, five years apart, same forum, same outcome. Frontier The most-cited scholarly reading, from Biermann and Gupta, is that African diplomatic leadership marks a paradigm shift with the Global South actively resisting Northern-led research agendas. Mark the interest: Frank Biermann initiated the Non-Use Agreement and Aarti Gupta is among its leading scholars. Their account of what states said is corroborated independently; the interpretive claim that this is a durable paradigm shift is advocacy-adjacent and carries the weaker flag.

Established While the international level failed twice, US states legislated — and these are now the only binding SRM prohibitions in force anywhere in the world. Tennessee enacted the first in April 2024, prohibiting the intentional injection, release or dispersion by any means of chemicals, chemical compounds, substances or apparatus within the borders of the state into the atmosphere with the express purpose of affecting temperature, weather or the intensity of sunlight. Florida followed with CS/CS/SB 56: filed 20 November 2024, passed the Senate 28–9 on 3 April 2025 and the House 82–28 on 30 April, approved 20 June and effective 1 July 2025 as Chapter 2025-157. It prohibits acts intended to affect temperature, weather or the intensity of sunlight, imposes a monthly reporting duty on public infrastructure operators — and repeals Florida's existing weather modification licensing framework outright. Frontier Nationally, per SRM360's tracking, 30 states have proposed such bills, 22 of them introducing proposals in 2025 alone; Arizona and Florida passed bills through their senates; and Illinois, Mississippi, New Hampshire, North Dakota, South Dakota, Utah and Wyoming rejected proposals after scientific and environmental review. The tracker is maintained by an SRM-focused non-profit and the underlying bills are public record.

Established Three observations follow, and they are uncomfortable for everyone. These statutes, not the CBD decision, not ENMOD, not the London Protocol amendment, are the binding law. They were passed largely on the strength of a false belief — the legislative energy behind them is substantially the belief that covert atmospheric spraying is already occurring, which is not supported by evidence and is answered at length in Climate Engineering. And in their prohibitory core they are roughly what the Non-Use Agreement campaign asks for. Frontier Mexico is the international analogue: after a startup's April 2023 release in Baja California the government announced an intention to ban all SRM experiments including research, though the available analysis does not establish that this became binding law, and its conclusion is the sharp one — the absence of proper regulations becomes a form of governance itself, one that benefits commercial projects outside democratic control.

Established Governing research turns out to be a distinct and harder problem than governing deployment, and it is failing in a specific way. Deployment governance is a coordination problem among states. Research governance is a consent problem between institutions and localities, and the record shows the latter is currently binding. Every framework specifies research governance in similar terms: the US National Academies would make codes of conduct a funding condition and add patent pledges; the US Office of Science and Technology Policy proposes responsible modelling, open data, adaptive oversight and public and Tribal consultation; ARIA operates the most elaborate scheme actually in force, with multi-stage approval, community engagement, environmental and legal assessment, an independent oversight committee and chief-executive sign-off; the EU review says outdoor experiments should require impact assessment and community consent. Established And yet the pattern is precise: no outdoor SRM experiment has been stopped by a regulator applying a rule. Every one has been stopped by a community objecting to a process. A Swedish balloon flight that would have released nothing, cancelled after the Saami Council objected; a sea-salt sprayer shut down by a unanimous city council vote over a lease and inadequate notice; three of ARIA's five outdoor projects still unable to name a site. The frameworks are not being enforced against experiments; the experiments are not reaching the point where a framework would apply.

Frontier Consent is functioning as the operative rule in the absence of law — effectively, but with no standard, no appeal and no reach. A researcher cannot know in advance what would satisfy it, and it operates only where the actor asks. Established There is a second-order failure the frameworks do not address at all: they are designed for publicly funded institutional research. In 2025 alone more than $60 million of commercial and philanthropic money entered the field, against roughly $170 million of cumulative public funding across two decades. A code of conduct enforced as a condition of a grant does not reach a venture-funded company with a proprietary particle.

Frontier Which forum, if any, could host this? Every candidate is imperfect and the imperfections are different. UNEA has failed twice. The CBD has a decision, a research carve-out and a non-party in the United States. The London Protocol has real regulatory machinery and a live process — over the marine environment only, with its own amendment at nine ratifications after thirteen years. The Montreal Protocol is the strongest environmental institution in existence, and Atmospheric Management's caution stands: its machinery is built around producers and substitutes, which SRM lacks, and loading it with new mandates risks the thing that already works. The UNFCCC has universality, no relevant machinery, and a negotiating culture that has never adopted a technology-specific prohibition. Martin and Moore propose the UNFCCC or Montreal; nobody has secured either. Speculative A dedicated multilateral treaty is the legal literature's preferred answer and the record is unkind to it: two UNEA failures, and thirteen years to nine ratifications on the one amendment the system did adopt.

Frontier Would any of it constrain a determined state? The literature disagrees and the disagreement is not resolvable from evidence. Martin and Moore, in the Harvard International Law Journal in 2025, argue that unilateral deployment presents a free-driver problem, that existing international law would provide only weak constraints, and — more alarmingly — that the jus ad bellum regime is unlikely to prevent unauthorised use of force against a unilateral actor, because a state perceiving another's deployment as a national security threat may resort to force and the law governing when force is justified would not reliably stop it. Horton argues the opposite from structure rather than law: interfering schemes, termination lock-in, and cheap countermeasures make cooperation rational, and using Underdal's typology he classifies SRM as a benign coordination problem where interests are relatively aligned, unlike mitigation or proliferation, making the standard nuclear analogy inapt. Speculative Horton's argument is the more interesting one and it has a hole: the countermeasure capability — fluorocarbon release, strategic black carbon — has never been modelled. An entire strategic-stability argument rests on a capability nobody has simulated.

Established And both address the wrong actor. Each argument concerns a state deploying at climate-altering scale. What has actually occurred is a two-person company with balloons selling cooling credits, and a venture-funded firm with an undisclosed particle. The theory was built for the wrong actor, which is why the London Protocol parties' 2025 expression of concern about commercial marine geoengineering is a better indicator of where this is going than any treaty proposal.

Frontier The research paradox has three horns, not two, and the third is the interesting one. The standard formulation is that governing research requires knowing its risks and knowing its risks requires research. Horn one is the non-use position: nothing at kilogram scale informs megatonne-scale behaviour, so research cannot reduce the decision-relevant uncertainty and functions only to normalise the option. Horn two is the counter-letter position: current knowledge is not sufficient to detect, attribute or project SRM consequences, which is itself a governance emergency, because without detection and attribution no state could tell whether another had deployed and no liability regime has anything to attach to — on this view refusing research does not prevent deployment, it prevents knowing about it. Speculative Horn three, which neither side emphasises: detection-and-attribution research and efficacy research have opposite governance profiles. Building the capability to detect a stratospheric perturbation and attribute an anomaly to it is a monitoring capability, requires no release, and is a precondition for enforcing any regime including a non-use one. Building the capability to know how well injection cools is a deployment-enabling capability. Both petitions bundle them as “SRM research.” Unbundling them is the most obviously available governance move that nobody has proposed as an instrument, and this pack found no framework that makes the distinction operational.

Frontier Definitional arbitrage between instruments is this slot's structural problem and it is under-noticed. Planetary Cooling Concepts records that cirrus cloud thinning sits in an awkward gap: it attracts the objections that attach to solar geoengineering, it is not solar geoengineering — it is a longwave lever — and it is routinely omitted from the frameworks being drafted for stratospheric aerosols. The parallel case runs the other way: marine cloud brightening, an unambiguously shortwave intervention, may come under a binding marine dumping treaty while stratospheric injection remains covered by nothing. Speculative A regime whose coverage is determined by which environmental medium a technique happens to touch is a regime that can be routed around by choosing a different medium, and no proposal in the literature addresses this.

Speculative And a fringe-but-persistent position deserves stating rather than ignoring: that governance is impossible and the only question is who deploys first. The evidence for it is the record in this brief — two UNEA failures, nine ratifications in thirteen years, no experiment governed by any rule. The evidence against it is that the London Protocol has actually regulated something, and that consent has stopped every institutional experiment attempted. Both are real, which is why the position is speculative rather than either established or dismissible.

Frontier The binding bottleneck is attribution, and it is logically prior to every other proposal in this brief. If a deployment could not be reliably detected and attributed for years, then treaty obligations, liability rules and countermeasure threats all lack a trigger. The signatories of the 2023 climate-intervention research letter name detection and attribution as an explicit research need, stating that current knowledge is not sufficient to detect, attribute or project the consequences of SRM interventions. It is the capability every position needs, including the position that deployment should never happen, and it is the one least funded.

Frontier The second is a definition that distinguishes research from deployment. The Make Sunsets analysis identifies this as lesson one: explicit, neutral definitions separating responsible from irresponsible experiments. The London Protocol's assessment framework is the only worked example of such a test anywhere in international law, and it applies to ocean fertilisation. Without such a definition, a moratorium cannot say what it prohibits and a permissive regime cannot say what it permits.

Established The third is reach over private actors. The ICJ's 2025 opinion articulates state obligations to regulate private conduct, and no state has exercised such regulation over an SRM actor except through the EPA's section 114 information demand and two state statutes. A startup with more than 124 deployments was answered with a request for information under a statute written for industrial air pollution in 1970.

Speculative The fourth is a liability regime, which requires the first. Nothing can attach to an unattributable act. And the fifth is legitimacy: UNEA-6 established that the populations with the most exposure have the least leverage in the institutions that might claim a mandate, and that the states best able to deploy resisted governance. Frontier No proposal solves this. The Degrees Initiative's capacity-building in the Global South is the main attempt and is itself contested — Geoengineering Monitor alleges that its staff and board sit in Global North institutions, that 80% of publication authors do too, that over 90% of funded studies model injection under the highest-emissions scenarios, and that three of six major funders also finance marine cloud brightening experiments. Interested parties on both sides: Degrees is a funder describing itself and Geoengineering Monitor campaigns against all geoengineering. The counts are largely undisputed and the interpretation is not. Declare the tie.

Established Nothing on this map produces the result this brief waits on, and the reason is structural rather than incidental. Climate Engineering supplies the facts this brief governs — the cost, the models, the outdoor record — but it produces no institution. Atmospheric Management holds the ozone regime and its finding about producers and substitutes, invoked here by reference and not restated. Planetary Stewardship asks whether any institution can manage a planet and answers with a two-success record. This brief assesses institutions; it does not wait on one being produced elsewhere on the map. What it waits on are facts about the world, recorded as typed requirements below, and this is the most institution-gated slot in the category.

Frontier One dependency is worth naming precisely because it is usually mislabelled as a research gap. Detection and attribution capability is not a discovery waiting to be made: the observing systems, the statistical methods and the relevant expertise all exist. What does not exist is a mandate — an institution charged with maintaining independent monitoring of atmospheric intervention and empowered to publish an attribution. That is an institutional fact, not a scientific one, which is why it is typed the way it is below.

Speculative And one dependency runs outside this map entirely. The strongest structural argument against unilateral deployment rests on counter-geoengineering, and no coupled-model study has ever simulated a deliberate counter-deployment. Whether the deterrent exists is a modelling question with a governance consequence, and it belongs to the climate engineering brief's experiment list rather than here — but every strategic-stability claim in this brief is conditional on it.

Established This slot's experiments are institutional, and the first one is already running whether anyone designs for it or not. The London Protocol's intersessional correspondence group reports to October 2026 on how existing instruments apply, on refined definitions for priority techniques, and on implementation of assessment frameworks — with marine cloud brightening among the techniques considered for Annex 4 listing. That is a live test of whether a functioning treaty body can extend a working framework to a solar geoengineering technique, and its outcome is the single most informative datum this subject will produce this decade.

Frontier Second, run a research-governance framework through to a completed outdoor experiment. No outdoor SRM experiment has ever been permitted, conducted and reviewed under one of the frameworks written for it. ARIA's approval architecture is the most elaborate in force and two of its five outdoor projects are active; whether a project completes under it, and what the oversight committee's review finds, would be the first empirical evidence that any of this machinery works.

Speculative Third, and cheapest: draft an instrument that separates monitoring research from efficacy research and see whether anyone will sign it. The distinction is coherent, requires no release, and is a precondition for enforcing a non-use agreement as much as for enforcing a permissive regime. Nobody has proposed it as an instrument, so its political feasibility is entirely unknown — which makes drafting it a cheap test of a real hypothesis.

Frontier Fourth, a serious study of what the state statutes actually do. Two prohibitions are in force and a third of US states have considered them; nobody has analysed their enforceability, their interaction with the Weather Modification Reporting Act, or the effect of Florida repealing a functioning cloud-seeding licensing regime. The world's only binding SRM law is unstudied, which is an odd place for a research field to leave a natural experiment. Speculative Fifth, a prospective study of whether a consent process can be designed to succeed: the field has three well-documented refusals and no documented success, and no methodology dedicated to the question.

Established The engineering requirement in this slot is an observing and attribution system, and it is the only piece of hardware every position needs. It would have to detect a stratospheric aerosol perturbation, distinguish it from volcanic and wildfire injections, and support attribution of regional climate anomalies — independently of any deploying party, and sustained for as long as the capability exists. Nothing about it is technically novel. Speculative It is unfunded, undesigned and unmandated, and the reason is that it belongs to no one: a research agency has no enforcement interest, an enforcement body has no research budget, and no treaty assigns the function.

Frontier The second requirement is a registry and assessment apparatus of the kind the London Protocol already built for one technique. An assessment framework is a piece of institutional engineering with a specification: a definition of the regulated activity, a test for what counts as legitimate research, a permitting authority, and a record. That machinery exists, works, and covers ocean fertilisation. Extending it is an engineering problem in the institutional sense — and thirteen years and nine ratifications is the measured throughput of the system that would have to do it.

Established The third is at the national level and is mostly repurposing. The instruments actually used against an SRM actor to date were a 1970s air-pollution statute and a 1972 weather-modification reporting requirement. Neither was designed for this, both had reach, and both produced information rather than prohibition. Domestic law that already exists is doing more work in this field than any instrument written for it.

Within this map: Climate Engineering, which supplies every fact this brief governs and takes none of the institutions; Carbon Capture at Scale, whose verification problem is the same problem in a different medium — and whose one regulated method, ocean fertilisation, is the worked example this brief's London Protocol section rests on; Atmospheric Management, which owns the ozone regime as a worked example of atmospheric governance and whose finding that it succeeded because the substances had identifiable producers, drop-in substitutes and legible harm is cited here as an input to the forum question rather than restated; Planetary Cooling Concepts, whose observation that cirrus thinning falls into a definitional gap is picked up here as a governance finding with the physics left there; Weather Modification, which owns the ENMOD-versus-cloud-seeding question and the older regulatory record; Ocean Engineering, from which this brief borrows only the structural parallel of a consensus-bound international body outpaced by national regimes; Space Law and Governance, the nearest analogue of a commons with capable private actors and thin instruments; and Planetary Stewardship, which asks the general question this slot asks specifically.

This brief supersedes the carbon programme's treatment as this slot's coverage; that page remains as the programme's shorter narrative entry. Outside the map: public international law, particularly transboundary harm and the law of treaties; the international relations literature on coordination and collective action; science and technology studies, which supplies the research-governance and consent literature; and US state legislative practice, which is where the only binding rules currently are.

Established Every expert body that has reported converges on the same narrow position, and the convergence is worth stating because it is real and because none of it binds. The US National Academies recommended a federal research programme of $100–200 million over five years, with international coordination and co-development from the outset, mandatory codes of conduct as a funding condition and patent pledges — supporting research only. The US OSTP's congressionally mandated report of 30 June 2023 sets five research goals and a governance framework of responsible modelling, open data standards and adaptive oversight with public and Tribal consultation, states that issuing the report signals no Executive Branch policy decision and that current research is not a preparatory measure for deployment, and adds that deployment would involve risks almost all of which are poorly understood and some unknown. Established It also states an aim to prepare the United States for possible deployment by other public or private actors, which is the first US government acknowledgment that the governing scenario may not be one it controls. The Climate Overshoot Commission recommends a moratorium on deployment and on large-scale outdoor experiments carrying transboundary-harm risk, while expanding research. The EU's Scientific Advice Mechanism presents options rather than recommendations — fund indoor research under strict governance, or abstain citing mitigation deterrence; require impact assessment and community consent for outdoor work; and it lists prohibiting deployment aimed at global temperature modification as one option among several. It does not recommend a moratorium.

Frontier The scholarly community is split, the split is substantive, and both petitions are worth counting rather than paraphrasing. The Solar Geoengineering Non-Use Agreement was launched in 2022 by 16 initiators with an open letter signed by academics from 40 or more countries, all in personal capacity and none institutional, asking governments to commit to five prohibitions: no public funding for development, no outdoor experiments, no patent rights, no deployment, and no support for SRM in international institutions. The Climate Intervention Research Letter of February 2023, signed by over 110 physical and biological scientists including James Hansen, asks for research into three specific things — the effectiveness of different interventions, their climate impacts under different scenarios, and the capabilities for detecting and attributing their impacts — under conditions of transparency, independence from political and commercial influence, Global South inclusion, and negligible environmental effect, stating explicitly that signing does not imply support for deployment. Frontier This is not research versus no research. It is a disagreement about whether research reduces or manufactures the probability of deployment, and it is not resolvable by available evidence because it turns on a counterfactual about institutional behaviour. Both are advocacy documents signed by serious people, and neither is a scientific finding.

Established The interested parties in this brief run in every direction and it is load-bearing to say which way each one runs. Biermann and Gupta's UNEA account comes from the Non-Use Agreement's initiator and one of its leading scholars, and is corroborated independently on the facts. CIEL campaigns for a non-use regime and the page cited does not reproduce the CBD decision text. SRM360 is an SRM-focused non-profit and its state-legislation tracking rests on public bills. ARIA describes its own programme. Degrees reports on itself and Geoengineering Monitor campaigns against everything Degrees funds. Frontier One source runs conspicuously against its own interest and is weighted up for it: the US government's own research plan, which states that current research is not preparation for deployment and in the same document sets out to prepare for deployment by others.

Frontier And the institutional requirement nobody has met is a forum with a mandate. Every candidate is either unable, unwilling, wrongly scoped or too valuable to risk. That is not a gap that better drafting closes; it is the reason two attempts at the mildest possible instrument failed in the only forum that tried.

Established The distribution of exposure and the distribution of leverage are inverse, and UNEA-6 demonstrated it rather than argued it. The states best able to deploy resisted governance; the states most likely to bear the consequences demanded it, invoking a continental ministerial call for a non-use mechanism. That is not a theoretical concern about procedural justice. It is the observed structure of the only negotiation the subject has had.

Frontier The most uncomfortable ethical fact in this brief is that the world's only binding SRM prohibitions were produced by a false belief. Tennessee and Florida prohibit approximately the conduct the Non-Use Agreement seeks to prohibit, and the legislative energy behind them was substantially the conviction that covert spraying is already occurring — a claim with no evidential support, answered in detail in Climate Engineering. Frontier Whether a correct rule reached for a wrong reason is a good outcome is a genuine question and this brief does not resolve it. What it records is the mechanism: rules in this field are being made by publics reacting to what they believe is happening, not by institutions responding to what is. One of those two processes is currently functioning.

Established Consent has been the operative ethic and it has worked in one direction only. Three well-documented refusals stopped institutional experiments; no comparable mechanism touched the commercial actor with 124 releases. A regime that constrains the transparent and not the opaque is the exact inverse of what governance is for, and it is what currently exists. Speculative It is also, on the evidence, the only functioning constraint in the field — which makes arguing against it awkward for anyone who wants a rule.

Frontier And legitimacy through capacity-building is contested by both interested parties. The Degrees Initiative reports five new country teams, fourteen renewals, roughly 60 papers, fourteen of its funded researchers contributing to the next IPCC assessment cycle, and a stated neutrality on whether SRM should ever be used. Geoengineering Monitor alleges Global North structural dominance and a funder-shaped research agenda. The underlying facts — paper counts, funder identities, board composition — are checkable and mostly not disputed, only interpreted differently. Declare the tie and record both.

Established The framing fails, and the record contains a clean natural experiment on it. Between 2019 and 2026 the international system attempted twice to establish the mildest possible instrument. UNEA-4 failed. UNEA-6 failed after the resolution had been compromised down to a proposal that UNEP prepare a paper on whether to establish an information repository — and that was still too divisive to adopt. Over the same period the one specific climate-intervention amendment the system did adopt accumulated nine acceptances in thirteen years and remains not in force.

Established In the same window, without any of that: a startup conducted more than 124 stratospheric sulfur releases and sold credits against them; another raised $75 million on a deployment-by-2035 slide with an undisclosed particle; two US states enacted the world's only binding prohibitions, largely on a false premise, one of them repealing a working cloud-seeding licensing regime to do it; and the federal response was an information request under a statute written for industrial air pollution. Deployment, at whatever scale, arrived first. So did the law — but sub-nationally, reactively, in one country, on a mistaken factual basis, and unconnected to any international instrument. The governance that exists was not built before the deployment; it was built because of it, and by people who thought something else entirely was happening.

Frontier The general pathology is worth naming, because it is not confined to this subject. Most planetary-scale capabilities are disciplined by their own cost: the institution required to govern them is cheap relative to the thing itself, so there is time to build it while the capability is assembled. Here the relationship inverts — the intervention costs less than the institution needed to govern it, and the institution is therefore never the path of least resistance for anyone. Every failure in this brief is an instance of that arithmetic rather than of bad faith or bad drafting.

Speculative The one hopeful reading is also the one that is hardest to build on. Consent works. The Saami Council, an Alameda city council and the communities ARIA's unsited projects have not persuaded have between them stopped every institutional outdoor experiment attempted. That is a functioning constraint and it is the only one. But it is not governance in any sense the framing means: no standard, no appeal, no reach over commercial actors, and it operates only where the actor asks. Handwave Whether a civilisation can convert an ad hoc veto into an instrument, before someone who does not ask reaches a scale that matters, is not a question the record answers, and any confident answer is assertion.

These horizons track treaty processes, legislative sessions and reporting dates rather than technology:

• 10 yr: Established One hard date dominates the near term: the London Protocol's intersessional correspondence group reports to October 2026, with marine cloud brightening among the techniques considered for listing — the only live route by which a solar geoengineering technique could come under a binding treaty. Frontier Expect the count of US state prohibitions to rise from two and the count of binding international instruments to stay at zero; the observed rate over 2024–2026 supports both. Speculative A third UNEA attempt is plausible and its failure mode would be informative either way. Speculative Expect no liability regime, because attribution capability is the precondition and nobody is funding it.
• 25 yr: Speculative If any binding instrument arrives in this window it most likely arrives through an existing body extending an existing framework — the London Protocol route — rather than through a purpose-built treaty, because that is the only mechanism in the record that has ever produced anything. Speculative The alternative and equally plausible path is that domestic law consolidates: a patchwork of national statutes, mutually inconsistent, with definitional arbitrage between them. Handwave Choosing between those two is asserting a political outcome.
• 50 yr: Speculative At this horizon the question is whether an intervention has been attempted at a scale requiring a response, and whether the response was legal or forcible. Martin and Moore's argument that the law governing use of force would not reliably prevent the latter is the most consequential claim in the legal literature and is untested. Handwave Everything else at this range is scenario writing.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The one durable observation is that a deployment begun and sustained across this horizon implies an unbroken institutional commitment longer than most constitutions have lasted, and that the instruments available today take thirteen years to reach nine ratifications. Handwave Nothing in the record supports a projection at this range, and stating one would be the failure this brief spends its length documenting.

• Depends on Nothing on this map. This brief assesses institutions; it does not wait on a result another brief produces. Climate Engineering supplies the facts to be governed — the cost, the model disagreement, the near-empty outdoor record — but produces no instrument; Atmospheric Management holds the ozone regime as a worked example; Planetary Stewardship asks the general form of the question. No typed depends-on edge is claimed.
• Requires (not on this map) Four things, all institutional, none of them a brief on this map — which is why this is the most institution-gated slot in the category. First, a forum with a mandate: UNEA has failed twice, most recently over a proposal that UNEP prepare a paper on whether to establish an information repository; the CBD has a decision with a research carve-out and a non-party in the United States; the London Protocol has real machinery over the marine environment only, and its own 2013 amendment has nine acceptances after thirteen years; the Montreal Protocol is built around producers and substitutes that solar radiation modification does not have, and loading it risks the strongest environmental institution in existence; the UNFCCC has universality and no relevant machinery. Second, a treaty definition separating legitimate research from deployment. The London Protocol's assessment framework for ocean fertilisation is the only worked example anywhere in international law, and without such a test a prohibition cannot say what it forbids and a permissive regime cannot say what it permits — which is also how a shortwave technique like marine cloud brightening can fall under a marine dumping convention while stratospheric injection falls under nothing. Third, an independent monitoring and attribution mandate. The observing systems and statistical methods exist; what does not exist is any body charged with maintaining independent detection of atmospheric intervention and empowered to publish an attribution, and without it no liability regime, no treaty obligation and no countermeasure threat has a trigger. Fourth, regulatory reach over a privately funded actor: the ICJ's July 2025 advisory opinion articulates state obligations to regulate private conduct as existing law, and the only exercises of such reach to date are a US EPA information demand under a 1970 air-pollution statute and two US state statutes — against research governance schemes that reach public grantees while more than $60 million of commercial and philanthropic money entered the field in 2025 alone.
• Enables Any outdoor experimental programme, any research agenda that requires a licence, and any deployment scenario in this category is gated on what is recorded here rather than on a discovery. No typed enabling edge is claimed, and the reason is the finding: no framework in this subject has ever been applied end to end to a completed experiment, so there is nothing yet for another brief to rest a result on.
• Adjacent Public international law, particularly transboundary harm and the law of treaties; the international relations literature on coordination and collective action; science and technology studies, which supplies the research-governance and consent literature; US state legislative practice, where the only binding rules currently are; and within this map Climate Engineering, Carbon Capture at Scale, Atmospheric Management, Weather Modification and Space Law and Governance.

Handwave “ENMOD covers geoengineering.” It does not, and this is the most confidently repeated error in the subject. ENMOD prohibits hostile use of environmental modification techniques having widespread, long-lasting or severe effects; a state cooling its own climate is not conducting hostile modification, the troika threshold is high and attempts to remove those qualifiers failed for want of consensus, and the convention has 78 parties. The treaty everyone names as the geoengineering treaty governs almost none of the conduct at issue.

Handwave “The CBD imposed a moratorium.” As a legal characterisation this is wrong. Decision X/33 is a COP decision under a framework convention containing an invitation to parties, with an explicit carve-out for small-scale controlled scientific research, reaffirmed in 2024, and the United States is not a party. Frontier As a normative claim the position is stronger: the decision does shape expectations and is cited by funders and researchers as a constraint. Two different claims travel under one word, and the stronger legal-sounding version is the false one.

Established “The London Protocol amendment bans marine geoengineering.” The 2013 amendment was adopted and has nine acceptances and is not in force. What continues to apply are the 2008 and 2010 resolutions, which are resolutions rather than treaty obligations. Frontier The more interesting correction runs the other way: the London Protocol is the most effective body in this field, not the least — it is the only one that has regulated a specific technique, it has a working assessment framework, and it has a live process that could bring marine cloud brightening under binding regulation. It is under-rated by the discourse relative to its actual output.

Established “The Non-Use Agreement has state signatories.” Its signatories page lists academics signing in personal capacity, from 40 or more countries, with no government endorsements. Established “The research community is uniformly pro-deployment.” Every institutional assessment — National Academies, OSTP, Overshoot Commission, EU Scientific Advice Mechanism — endorses research while declining to endorse deployment, and the letter of 110-plus scientists states explicitly that signing does not imply support for use. Frontier “Opposing a moratorium means favouring deployment.” A serious position holds that a moratorium would be unenforceable, would push work into less transparent settings, and should be decided by a legitimate representative body rather than declared. That is an argument about instrument choice, not about outcomes.

Frontier “Research governance frameworks are working.” They have never been tested: no outdoor SRM experiment has been permitted, conducted and reviewed under one. The frameworks are elaborate and in force and the experiments are not reaching the point where they would apply. Established And the related mis-description: “governance is being outrun by activity.” The record is more specific and more interesting than that. No experiment has been stopped by a regulator applying a rule; every one has been stopped by a community objecting to a process. That is a different diagnosis with different implications, and it means the constraint that binds is one no framework has written down.

Speculative “Counter-geoengineering would deter a unilateral deployer.” Named in the strategic literature — fluorocarbon release, strategic black carbon, a competing scheme in the opposite hemisphere — and never simulated in a coupled model. Frontier “Existing law would constrain a determined state.” Martin and Moore say it would provide only weak constraints and that the law on use of force would not reliably prevent an armed response; Horton says structural factors would constrain, which is a different claim with no test behind it. Neither should be presented as the settled reading.

Handwave “Covert large-scale deployment is already occurring, so governance is moot.” The claim is false and is answered with the evidence in Climate Engineering: 76 of 77 surveyed domain experts find no evidence, the required high-altitude tanker fleet does not exist, and the stratospheric aerosol record — one of the best-monitored quantities in Earth observation — does not show it. Frontier It is nonetheless recorded here rather than omitted, because it is the position that has produced the most enacted law in this field. That is a fact about how rules actually get made, not a vindication of the belief, and it is the single strongest piece of evidence that this subject's governance is being driven by publics rather than by institutions.

Frontier “Capacity-building has resolved the legitimacy problem.” One funder reports 60 papers and teams in fourteen or more countries and states neutrality on use; a campaigning organisation alleges Global North structural dominance and a funder-shaped agenda. Both are interested; the counts are largely undisputed and the interpretation is not. Declare the tie. Established And the framing itself: “the governance can be built before the deployment.” On the evidence of the last seven years it was not. It was built afterwards, at the wrong level of government, in one country, on a mistaken premise — and the only mechanism that has actually stopped anything has no standard, no appeal and no reach over the actors that do not ask.