CXC Rewrote the SBA to Stop AI Cheating. Teachers Say That Solves Only Half the Problem

Two things happened in Caribbean education nine days ago, and they were pointed in almost opposite directions. On July 16, the Caribbean Examinations Council announced it is rewriting how most CSEC and CAPE candidates are assessed, pulling the traditional School-Based Assessment out of students' own time and back into a supervised exam room. On July 18, the Jamaica Teachers' Association responded by asking for something CXC's fix cannot give it: a real national policy for how AI gets used, and taught, everywhere a student sits with a screen. Read together, they describe an exam board closing one door while a teachers' union asks who is going to open the right one.

Both moves are honest reactions to a real problem. Neither one, on its own, teaches a Caribbean teenager the one skill the region's own labour data says it needs most: how to build something with AI, not just how to avoid getting caught misusing it.

What CXC Actually Rewrote

CXC's Registrar and CEO, Dr Wayne Wesley, framed the reform in a single line: "The integrity of our qualifications is not negotiable." Director of Operations Dr Nicole Manning put the stakes in terms parents will recognise: "A CXC qualification means something. It means something to employers, to universities, to parents, families and guardians, who have invested years of commitment and sacrifice into a child's education." That is the problem the reform is aimed at, not classroom pedagogy but the basic promise that a grade reflects the student who earned it.

Practical subjects, Agricultural Science, Visual Arts, Music, Physical Education, Technical Drawing, and Food, Nutrition and Health, keep their traditional, project-based SBA, just under tighter moderation. Everywhere else, Mathematics, English, Caribbean History, Social Studies, Principles of Business, and most non-practical subjects, the SBA is being replaced by Paper 032, an existing alternative assessment CXC is now expanding regionwide. Instead of weeks of take-home coursework, candidates sit a supervised paper, with three built-in concessions: the topic is shared roughly a month ahead, candidates get extra time to complete it, and reference notes are allowed in the room. CAPE candidates move fully to Paper 032 from the May-June 2027 sitting. CSEC candidates get a choice in 2027, with Paper 032 becoming mandatory by May-June 2028. The change followed consultations across 21 Caribbean states and territories, which is a genuinely regional process, not a decision handed down from one office.

Read on its own terms, this is a sensible fix for a real fraud problem. A take-home assignment that anyone can quietly hand to a chatbot stops proving what it was designed to prove. Moving the assessment into a room where nobody can do that is a direct, workable answer. What it is not is a plan for what a student does with AI the other 179 days of a school year that are not exam week.

The Part That's Actually New: Teachers Asking for the Whole Policy

This is where the story stops being only about exam mechanics. The JTA's response, filed two days after CXC's announcement, did not just react to Paper 032. It named the actual scope of the problem: AI misuse reaches into homework, internal school exams, and research assignments too, well past anything CXC's authority covers. The association's own words were direct: "Schools cannot be left to navigate these complex ethical, instructional and assessment issues individually." Its ask was specific, an urgent convening of teachers, school leaders, parents, students, universities, and technology experts to build a genuine national AI-in-schools policy, not a patch bolted onto one exam paper.

The equity point buried in that call deserves more attention than it has gotten. The JTA flagged unequal access, to devices, to reliable internet, to AI platforms themselves, and to teachers trained well enough to guide any of it, as a real risk that a policy vacuum could deepen. A student at a well-resourced school with fast broadband and a teacher fluent in AI tools is having a different education right now than a student at a rural school without either. Paper 032 treats every candidate the same way in an exam room. It does nothing about the gap in what happened before either of them walked in.

The Gap the Region's Own Data Keeps Pointing At

Step back from the exam calendar and the numbers explain why this gap matters more than a policy footnote. The Eastern Caribbean Telecommunications Authority's Digital Inclusion Survey, launched in January 2026 across Dominica, Grenada, Saint Kitts and Nevis, and Saint Lucia, found that broadband connectivity has grown sharply since 2014. Getting online is largely solved. What is not solved: roughly 60 percent of users across those member states have only basic digital skills, and fewer than 5 percent reach intermediate or advanced levels. ECTEL's own framing is blunt, the region has moved from a connectivity problem to a capability problem. Most people can reach the internet. Far fewer can build anything meaningful once they are there.

That capability gap lands directly on the job market. The ILO's report, A Generation in Transition: Rethinking Youth Employment in the Caribbean, put regional youth unemployment, excluding Haiti, at 17.6 percent in 2024, against 4.7 percent for adults aged 25 and over. Including Haiti, that figure rises to 30.5 percent. Young women face unemployment 1.6 times higher than young men. And the share of youth not in employment, education, or training, 24 percent excluding Haiti, reversed a years-long decline in 2024 and is expected to keep climbing. A generation facing those odds cannot afford a school system whose main AI policy achievement is teaching students what not to do during a two-hour paper.

Put the two data sets side by side and the shape of the actual problem gets clear. Connectivity is mostly solved. Integrity rules are getting built, sensibly, subject by subject. What is missing is the middle piece: structured time for a Caribbean teenager to actually build something with AI, get it wrong, fix it, and build it again, which is the only process that turns basic digital use into the intermediate or advanced skill ECTEL found almost nobody has.

What Building Actually Looks Like

One Caribbean programme already shows what the missing half of this policy conversation looks like in practice. STEM SVG, run in Saint Vincent and the Grenadines by the National Centre of Technological Innovation, ran a three-week intensive from July 13 to 31, 2026, at St Martin's Secondary School, with more than 60 primary and secondary students choosing classes in Web Design, App Development, Graphic Design, Music-making, and core sciences, every one of them built around AI applications. Founder Petrus Gumbs was clear about the goal: AI is "a tool that can be harnessed to benefit themselves tremendously, academically as well," not simply entertainment, and not simply a threat to be policed.

That is the same instinct The Genius Project was built on. During a recent summer cohort session, one of our students, working through her first real AI-assisted build, put the difference between the two kinds of policy plainer than any of the official statements: "Dem a guh test mi pon paper, but nobody a teach mi how fi build the ting. If mi can build it, mi nah fraid a no exam." Nobody was teaching her how to build it. If she can build it, she is not afraid of any exam. That is the whole argument in two sentences, and it is exactly the gap between Paper 032's integrity fix and a full national policy that actually develops a skill.

What This Means for a Caribbean Teen Right Now

None of this is a reason to dismiss what CXC or the JTA have done this month. Paper 032 solves a real fraud problem carefully, with genuine regional consultation behind it. A national AI-in-schools policy, built the way the JTA is asking for, with teachers, universities, parents, and students at the table together, would be a real step forward for equity and clarity. Both are worth taking seriously on their own terms.

What a student heading into a new school year should take from this is narrower and more useful: whatever policy eventually gets written will tell you what you cannot do with AI in an exam room. It is very unlikely to teach you what you can build with it everywhere else. That second half is not going to arrive from a ministry memo on the same timeline as Paper 032. It has to be built deliberately, through practice, feedback, and real projects, the same way STEM SVG is building it in Saint Vincent this July and the way The Genius Project builds it with Caribbean youth year-round, tuition-free.

StarApple AI, recognised as the first AI company built in the Caribbean, has argued for years that the region's advantage will come from teaching young people to use AI transparently and skilfully, not from writing longer lists of what they cannot do with it. Its founder, Adrian Dunkley, has made that case publicly as the region's most recognised AI voice, and it lines up with what the data in this article actually shows: a connectivity problem the region has mostly solved, a capability gap it has barely started closing, and a labour market that is already sorting young people by which side of that gap they land on. CXC's Paper 032 and the JTA's policy call are both aimed at the right general problem. The part of it that changes a Caribbean teenager's actual future is the half neither one is built to cover.