The International Seabed Authority is holding the second half of its 31st annual session in Kingston from 29 June to 31 July 2026, deciding the rules for mining manganese, nickel and cobalt from international waters. Modern sonar has mapped 28.7 percent of the world's seafloor. Less than 5 percent has been seen with a camera. NOAA puts direct seabed sampling at roughly 0.001 percent.
Jamaica is hosting a vote on ground nobody has actually looked at.
I have spent the past two decades building AI capacity across this region, and I have watched enough policy fights to know when the argument on the table is really an argument about missing data. This is one of those fights. The people asking for a pause are not asking for it because they oppose mining on principle. They are asking for it because the regulator writing the rulebook does not have a map of the room it is regulating. That is a solvable problem, and it is solvable faster than the diplomatic calendar suggests, because the tools that close a knowledge gap this size did not exist five years ago and now sit inside a handful of ocean research labs a plane ride from where the ISA is meeting.
What's Actually on the Table in Kingston
The ISA's 31st session runs in two parts. The Legal and Technical Commission met in March, and the Council reconvened at ISA headquarters in Kingston from late June through the end of July to advance work on the mining code, the framework of standards and guidelines that would govern commercial seabed extraction once, or if, it is permitted to begin. The stakes are not abstract for the 170 member states in the room, 169 countries plus the European Union, because whatever code emerges from this session becomes the operating rules for every future licence application.
On 21 July, environmental groups renewed their push for Jamaica's government to back a moratorium during the session. Dr. Theresa Rodriguez-Moodie, chief executive of the Jamaica Environment Trust, put the argument in a single line: "We know more about outer space than we know about the deep sea." Two days earlier, on Sunday 19 July, activists staged a paddle-out at Jamnesia Surf Camp in Bull Bay, St Andrew, one stop in a coordinated global day of protest against seabed mining. Matthew Gianni, co-founder of the Deep Sea Conservation Coalition, and Davia Hylton, co-coordinator of the Stronger Caribbean Together Network, backed the call publicly. More than 40 governments, including Mexico, Brazil and the United Kingdom, now support a moratorium, a precautionary pause, or an outright ban on mining the international seabed.
None of that stopped the clock on the applications already in front of regulators elsewhere, which is the part of this story that makes Kingston's session harder than a normal environmental debate.
The Knowledge Gap Driving the Standoff
Strip the diplomacy away and the disagreement in Kingston comes down to three numbers. Modern multibeam sonar, the kind that produces a usable bathymetric map rather than a rough depth reading, has covered 28.7 percent of the seafloor. Visual exploration, meaning an actual camera or a human eye on the bottom, has reached less than 5 percent. NOAA's own estimate for direct seabed sampling and analysis sits near 0.001 percent, a figure so small it is easy to misread as a typo the first time you see it written down.
Those numbers matter because a mining code has to specify what counts as unacceptable environmental harm, and you cannot define harm to a system you have not measured. Every polymetallic nodule field under consideration for extraction sits on top of an ecosystem that took millions of years to form, hosting species that in many cases have never been catalogued, let alone assessed for how a mining plume, the cloud of disturbed sediment a collector vehicle kicks up as it moves across the bottom, would affect them. Regulators writing thresholds for that plume are, in a literal sense, guessing at a baseline they do not have.
This is not a new observation and it is not unique to this session. It is the same argument that has stalled the mining code for years. What is different in 2026 is that the tools for closing the gap have moved from research papers into operating projects, which changes what "we need more time to study this" should actually mean at the negotiating table.
Why the United States Walked Around the Table
The moratorium debate in Kingston carries less weight than it should, because the country most likely to start commercial mining first is not fully inside the room. The United States never ratified the UN Convention on the Law of the Sea, so it does not recognise the ISA's exclusive jurisdiction over the international seabed the way its 170 members do.
President Trump signed an executive order on 24 April 2025, titled Unleashing America's Offshore Critical Minerals and Resources, directing federal agencies to fast-track domestic permits for seabed mineral extraction using a 1980 law, the Deep Seabed Hard Mineral Resources Act, that predates the modern ISA framework entirely. NOAA finalised a rule on 21 January 2026 allowing companies to apply for an exploration licence and a commercial recovery permit in a single consolidated process instead of the traditional two-step sequence. The next day, 22 January 2026, The Metals Company USA filed the first such application, covering 65,000 square kilometres of the Clarion-Clipperton Zone, a stretch of the Pacific between Hawaii and Mexico that every ISA member treats as international territory. NOAA determined the application was in substantial compliance with US law that March.
Nothing about that sequence required a vote in Kingston. It ran entirely through American domestic regulation, using a legal theory that most of the ISA's own members reject. That is the second knowledge gap in this story, not about the ocean floor but about jurisdiction, and it is the reason environmental groups describe the moratorium push as racing a clock that is already running somewhere else.
Why This Isn't Just an Environmental Story
A mining code that gets finished in Kingston without a real seafloor baseline will be weaker on every axis: harder to enforce, easier to challenge legally, and less useful to the Caribbean states that have the least capacity to dispute a mining company's own environmental data. Closing the knowledge gap is not a favour to conservationists. It is the difference between a rulebook that holds up and one that gets tested the first time a contractor disputes a finding nobody can independently verify.
What AI Actually Changes About This Fight
Ocean mapping used to mean chartering a research vessel, towing a camera sled across a survey line for weeks, and paying a team of biologists months of review time to classify what the footage caught. That cost structure is precisely why 95 percent of the seafloor has never been seen. It was never going to get mapped at that pace before the mining applications arrived.
AI changes the arithmetic on both ends of that process. Plymouth Marine Laboratory and the University of Plymouth launched the Deep Vision project in March 2026, funded through the Bezos Earth Fund's AI for Climate and Nature Grand Challenge, to build the first comprehensive AI-generated maps of vulnerable marine ecosystems across the Atlantic. The models are trained to detect and classify indicator taxa, deep-sea corals and sponges among them, directly from towed-camera and remotely operated vehicle footage, turning a review process that used to consume months of a specialist's time into an automated pass that flags the frames worth a human's attention. NOAA Ocean Exploration is running its own 28-day mapping expedition in July and August 2026, using remotely operated vehicles to characterise deep seabed habitats it has never surveyed before.
The mapping side benefits from a parallel advance. Researchers have documented transferable satellite-derived bathymetry, a machine learning approach that extends limited sonar survey data into broader depth estimates using satellite imagery, rather than requiring a ship to physically cross every square kilometre it wants to chart. That does not replace direct sonar and sampling, and nobody serious claims it does. It does mean a region without a standing research fleet can generate a usable first-pass map far faster than the traditional survey-line approach, which is exactly the capability gap keeping the Caribbean and most of the Global South out of this data conversation entirely.
There is a third piece, and it matters more once any code or moratorium actually gets adopted: enforcement. A mining code is only as good as the regulator's ability to verify a contractor is operating inside its licensed coordinates. AI-based maritime monitoring, cross-referencing vessel transponder data and satellite imagery against a registry of licensed exploration blocks, can flag a support vessel the moment its track deviates from its permit area. That is the same evidentiary approach maritime domain awareness platforms already apply to track fishing fleets and drug-interdiction operations at sea. Applied to seabed mining, it turns compliance from a self-reported filing into a standing, independently verifiable record, which is the single thing a Caribbean state with no navy of its own most needs from a global mining regime it cannot police by other means.
The Opportunity Already Sitting in Kingston
Jamaica has hosted the ISA since 16 November 1994, the day the UN Convention on the Law of the Sea entered into force. That is 32 years of a standing diplomatic and scientific convening role that almost no other small state gets handed by accident of treaty history. The ISA already runs a capacity-building initiative called Advancing Caribbean Blue Economy Through Deep-Sea Research, aimed at strengthening regional research capacity and training Caribbean women scientists in deep-sea work specifically.
That infrastructure exists. What it is missing is the AI layer that would let it produce something the mining code negotiators can actually use instead of a long-term training pipeline measured in careers rather than sessions. UWI's Discovery Bay Marine Laboratory, on Jamaica's north coast, is the kind of institution built to anchor that work: an existing marine research station with the field access and the scientific relationships a mapping and monitoring programme needs, sitting a few hours from the ISA's own front door.
I am not going to pretend Jamaica can single-handedly fund an ocean-AI mapping programme that closes a 28.7 percent coverage gap on its own budget, because it cannot, and neither could most ISA members acting alone. What Jamaica can do is what it has not yet done: use its position as host nation to convene the ISA, UWI, and the AI ocean-mapping labs already doing this work elsewhere, and push for the mining code to require AI-accelerated baseline mapping as a funded condition of any future licence, not an unfunded aspiration in the preamble. That is a Kingston-shaped opportunity, built on an institution already physically present, and it is being left on the table while the meetings happen a few miles from where I am writing this.
What Has to Happen Before 31 July
The session closes at the end of the month, and closing without a concrete data commitment means the same argument reconvenes at the next Council meeting with the same 28.7 percent sonar coverage on the table. Two things would change that.
First, the mining code needs a funded, time-bound baseline mapping requirement built on AI-accelerated methods, not a promise to keep studying the issue. A requirement that named which portions of any proposed mining block had been AI-mapped and classified before extraction, with a public deadline attached, would give the moratorium camp and the mining applicants a shared standard to argue about instead of a shared absence of data to argue around.
Second, whatever compliance mechanism the code adopts needs an independent monitoring layer that does not rely on contractor self-reporting. An AI-based vessel and licence-area cross-reference system, run through or alongside the ISA rather than through any single member state, would give every government in that room, including the smallest ones with no independent enforcement capacity of their own, a way to verify what is actually happening on a mining block without needing a navy or a satellite fleet of their own.
Neither of those fixes settles the deeper argument, and I want to be honest about that rather than pretend a mapping tool resolves a political fight. Even a fully mapped, fully classified seafloor does not answer whether commercial extraction should happen at all. That is a values question about how much risk a fragile, barely understood ecosystem should carry for a supply of critical minerals the world could, in many cases, source or recycle differently. AI removes the excuse that nobody has the data to make that decision honestly. It does not make the decision for anyone, and the countries meeting in Kingston this month still have to make it.
The Caribbean's Stake in Getting This Right
The Caribbean is not a bystander in this fight, whatever the geography of the Clarion-Clipperton Zone might suggest. The region's tourism economy, its fisheries, and its own coastal ecosystems depend on ocean health in ways that a mining decision made about a Pacific seabed can still touch, through precedent if nothing else. A mining code finished in Kingston with weak data standards becomes the template every future application points to, wherever in the world it is filed next.
Jamaica did not choose to host the body making this decision. The 1994 treaty made that choice. What Jamaica chooses now, in the closing days of a session already running in its own capital, is whether to treat that as a diplomatic obligation to manage or a scientific and technological opportunity to build. I know which one I would choose, and I know which one leaves this country with something to show for the next ISA session beyond another set of meeting minutes.
Frequently Asked Questions
What is happening at the International Seabed Authority's 31st session in Kingston?
The International Seabed Authority is holding the second part of its 31st annual session at its Kingston, Jamaica headquarters from 29 June to 31 July 2026. Delegates from the ISA's 170 members, 169 states plus the European Union, are negotiating the mining code that would govern commercial extraction of manganese, nickel, cobalt and rare earth elements from international waters, and weighing whether to adopt a moratorium or precautionary pause while that code remains unfinished.
Why are environmental groups calling for a deep-sea mining moratorium?
Groups including the Jamaica Environment Trust, the Deep Sea Conservation Coalition and the Stronger Caribbean Together Network argue that regulators cannot write safe mining rules for an environment nobody has properly studied. Jamaica Environment Trust CEO Dr. Theresa Rodriguez-Moodie has said the world knows more about outer space than about the deep sea. More than 40 governments, including Mexico, Brazil and the United Kingdom, now support a moratorium, a precautionary pause, or an outright ban.
How much of the ocean floor has actually been mapped or explored?
Modern sonar has mapped 28.7 percent of the global seafloor. Less than 5 percent has been visually explored with cameras or remotely operated vehicles. NOAA estimates that roughly 0.001 percent of the deep seabed has been directly sampled and analysed. Nearly all of the ecosystems that would be affected by commercial mining have never been observed by a human being or a camera.
Why is the United States moving ahead with deep-sea mining outside the ISA process?
The United States never ratified the UN Convention on the Law of the Sea, so it does not recognise the ISA's exclusive jurisdiction over the international seabed. President Trump signed an executive order on 24 April 2025 directing federal agencies to fast-track domestic permits for offshore mineral extraction under the 1980 Deep Seabed Hard Mineral Resources Act. NOAA finalised a rule on 21 January 2026 that lets companies apply for exploration and commercial recovery permits at the same time, and on 22 January 2026 The Metals Company USA filed a consolidated application covering 65,000 square kilometres of the Clarion-Clipperton Zone in the Pacific, an area every other government treats as ISA territory.
How could AI help resolve the deep-sea mining knowledge gap?
AI models trained on sonar and camera imagery can classify seafloor features and identify vulnerable marine ecosystem indicator species such as deep-sea corals and sponges far faster than manual video review. The Deep Vision project, launched by Plymouth Marine Laboratory and the University of Plymouth in March 2026 with Bezos Earth Fund funding, is applying this approach to map vulnerable ecosystems across the Atlantic. Machine learning models can also extend limited sonar survey data into broader depth estimates using satellite imagery, a technique documented as transferable satellite-derived bathymetry, cutting the ship-time cost of building an actual baseline of what is down there.
Can AI actually stop illegal or unauthorised deep-sea mining?
AI cannot stop a licensed vessel from mining inside its own permit area, and it cannot force a country outside the ISA framework to comply with a regional moratorium. What it can do is give regulators an independent, continuous picture, cross-referencing vessel transponder data and satellite imagery against licensed exploration blocks to flag when a support vessel operates outside its permitted coordinates. That turns enforcement from an occasional inspection into a standing, verifiable record, the same evidentiary approach maritime domain awareness systems already use to track fishing fleets.
What does this mean for Jamaica and the wider Caribbean economically?
The International Seabed Authority has been headquartered in Kingston since 16 November 1994, giving Jamaica a permanent diplomatic and scientific convening role most small states never get. The ISA already runs capacity-building work under its Advancing Caribbean Blue Economy Through Deep-Sea Research programme. Pairing that existing presence with AI-trained ocean scientists, anchored at an institution such as UWI's Discovery Bay Marine Laboratory, would let Jamaica supply the mapping and monitoring expertise the entire mining code negotiation is short of, rather than simply hosting the meetings where other countries argue about it.
What would it take for the Kingston session to produce something the Caribbean can actually use?
The session needs to leave Kingston with a funded, time-bound baseline mapping requirement built on AI-accelerated survey methods rather than an open-ended promise to study the issue further. It also needs a monitoring mechanism that does not depend on voluntary disclosure from mining contractors. Short of that, the moratorium debate repeats at the next session with the same 28.7 percent sonar coverage and the same argument about who has to prove what before extraction can begin.