Five Mistakes Caribbean Parents Are Making With AI This School Year
Schools reopen across the Caribbean this week and next, and every child walking through the gate carries the same class of AI that NASA is building into its next space station. Are they using it to cheat on homework, or to learn faster than any generation before them?
NASA has been building a ChatGPT-style assistant for astronauts on the Lunar Gateway station since 2023, per NASA engineers quoted at the time.
It is 9pm. The homework is finished faster than ever, you are relieved, and something about it bothers you. Both feelings are right: the same tool can do a child's thinking or multiply it. Pew found 64% of teenagers use AI chatbots and only 51% of parents know. The five sections below cover the choices that tip it one way or the other, each with a picture you can read in a minute.
Mistake 1Assuming your child is not using AI when Pew says 64% of teenagers are
Pew Research Center surveyed 1,458 American teenagers and their parents in late 2025 and published in February 2026. Sixty-four percent of teenagers use AI chatbots; 51% of their own parents believe they do. Three in ten teenagers use one daily. Homework help tops the list of uses, and 12% use chatbots for emotional support.
The Caribbean version of the gap is probably wider, because of where the AI sits. An American teenager opens a dedicated app. A Jamaican, Trinidadian, Bajan, or Guyanese child meets Meta AI inside WhatsApp, the app the household already runs for school groups, family news, and the church bulletin. The second regional factor is the shared device: a parent's phone or a school laptop several students log into, where the account belongs to an adult and every teen protection described in Mistake 4 is off.
Ask the two questions the American Psychological Association recommended in February 2026: "How are you using AI?" and "How are your friends using it?" Ask, do not lecture; a child expecting a lecture says "I don't". Then look at WhatsApp together.
Mistake 2Choosing between a total ban and a free pass while CXC replaces the SBA
Caribbean parents tend to pick a camp. The first bans chatbots for schoolwork under any circumstances. The second treats the chatbot as a homework appliance and stops asking once the grades hold. Both camps run into the same institution this term, because the Caribbean Examinations Council has changed the rules twice in twelve months and most households have absorbed neither change.
In August 2025 CXC announced that from 2026 candidates could use AI in School-Based Assessments for CSEC, CAPE, and CCSLC under a published document, Standards and Guidelines for the Use of Artificial Intelligence in CXC's Assessment. Director of Operations Dr Nicole Manning told the Cayman results ceremony that "there is a standard that students need to abide by". A memo of 10 March 2026 required schools to run submissions through an approved AI checker, with CXC moderators making the final call. Then in July 2026, after consultations across 21 territories, CXC announced that a supervised Paper 032 will replace the SBA in non-practical subjects: CAPE from May 2027, CSEC optional in 2027 and mandatory in 2028. In the Newsletter's reading, CXC could no longer tell which SBAs a student had written, and moved the assessment into a room where it can.
The free-pass camp is training a child for a format CXC has abolished. The total-ban camp is producing a child who uses the tools in secret, untaught in the disclosure rules CXC applies this year, and one originality report away from a penalty that can move a subject grade across a boundary. CXC Registrar Dr Wayne Wesley has tied the whole policy to a blunter number: only 36% of candidates passed CSEC Mathematics in 2024, and a chatbot cannot repair that for a child who cannot check its arithmetic.
Download CXC's AI standards from cxc.org and read them with the child; they are short. Then set the household rule that mirrors the exam rule: AI may be used if the child can say what they asked, what came back, and what they changed. If the child cannot explain the work with the chatbot closed, the work is not theirs yet.
Mistake 3Missing the difference between a study tool and a companion app
ChatGPT, Gemini, Claude, and Copilot are general assistants. Character.AI, Replika, and dozens of similar products are companion apps, built to sustain a relationship with a character the user designs. A parent who has heard of the first group and never of the second is missing the category that produced the lawsuits.
Common Sense Media's 2025 study found 72% of American teenagers had used an AI companion and 52% used one a few times a month or more; it rated companions an unacceptable risk for anyone under 18. A Drexel University study in April 2026 read more than 300 Reddit posts by 13 to 17 year olds about Character.AI and found teenagers describing withdrawal, relapse, and emotional conflict in their own words. Wrongful death lawsuits have been filed against both OpenAI and Character.AI by parents of children who died by suicide after months of confiding in a chatbot.
Adolescent mental health services in the Caribbean are thin; a teenager in a rural parish, an outer island, or the Guyanese interior may have no counsellor at all, and a chatbot that is available at 2am, never judges, and always agrees is a plausible substitute when nothing else is on offer. The Child Mind Institute's March 2026 guidance is that chatbots deepen isolation when they become the main source of support. Six signs that a chatbot has moved from tool to companion:
Two or three signs together are a reason to ask, calmly, what the app is and who the character is, and to sit in the app beside the child before deciding anything. Cutting off an attached child abruptly tends to produce the withdrawal Drexel describes; a phased reduction with a conversation alongside it works better. Any mention of self-harm goes to a clinician or the school the same day.
Mistake 4Leaving ChatGPT for Teens and its parental controls switched off
For most of the past three years there was nothing to switch on. OpenAI launched parental controls in late September 2025 after a wrongful death lawsuit and a US Senate hearing, expanded the safety notifications in July 2026, and on 18 August 2026 began rolling out ChatGPT for Teens worldwide: a separate experience applied to any account that states an age under 18 or that OpenAI's age-estimation judges to be a minor, with stricter rules on self-harm, sexual content, and dangerous activity.
The controls are worth switching on, with the limits OpenAI itself states. Linking gives no access to chat history, and Common Sense Media's Robbie Torney has said the company still needs to show evidence that the protections work. The Caribbean complication is the shared phone again: a child using a parent's account, or one created with a false birth year, is in the adult experience with no safeguards and no link. OpenAI's age-estimation can move such accounts on behavioural signals, but the company does not present that as a guarantee.
Create the child's own account, link it, set quiet hours that match bedtime, and do the same in Google Family Link for Gemini. Then tell the child what you switched on and what you cannot see. The APA's finding is that rules made with the teenager hold better than rules handed down, and a control the child knows about is one they are less likely to route around with a second account.
Mistake 5Never teaching the child to check what the chatbot said
The mistake that outlasts the others is treating chatbot output as an answer rather than a draft. Every major model produces fluent statements that are wrong, at a measurable rate that rises on Caribbean subjects. In the Newsletter's testing across 2026, general-purpose models have misdated regional events, invented quotations from Caribbean figures, confused the constitutions of different territories, and produced Patois with words borrowed from other Creoles. A History SBA built on an unchecked summary of the Morant Bay Rebellion is built on whatever the model guessed.
The same habit protects the household in the other direction. Voice-cloning scams that impersonate a relative in distress, deepfake images of classmates, and fabricated screenshots are circulating in the region, and children are both targets and, sometimes, producers. The household defence is a spoken family password and a rule that no image of a person travels onward until its origin is known.
Pick one piece of homework the child used AI on this week and check three facts from it together against a non-AI source. Agree the family password the same evening. Put both into the agreement below so they survive past the first week of term.
A chatbot is a multiplier on what the child already knows. The parent's job this school year is to make sure the number it is multiplying is above zero, and that the child can tell when the product is wrong.
- Which tools. The study tools the household allows, named. Companion apps are not on the list.
- Which accounts. The child's own account, real name and real age, linked to a parent. No shared adult accounts for schoolwork.
- The disclosure rule. The child can say what they asked, what came back, and what they changed. This is the CXC rule at home.
- The checking rule. One non-AI source for any chatbot fact that goes into schoolwork, and the source written down.
- Hours. Quiet hours set in the controls and matched to bedtime, with the parent saying which and why.
- What the parent can and cannot see. The parent can set limits and receive safety alerts, cannot read conversations, and says so to the child in those words.
- The family password. One spoken word for any caller claiming to be family in trouble; no image of a person forwarded until its origin is known.
- The review date. Reread at half-term and in January, and changed if the child has earned more room or shown a need for less.
Frequently asked questions
Every child starting school in the Caribbean this week will sit exams under rules written for a world where the chatbot exists. The parents who do best this year will be the ones who ask, every week, what the child can explain with the chatbot closed.Caribbean AI Newsletter · The Classroom Shift, 1 September 2026
CXC Just Set the Rules for AI in Your SBA. Here Is What Actually Changed
The Genius Project's plain-language guide to CXC's May 2026 SBA policy: disclosure forms, originality reports, and what counts as dishonesty.
Read the guide →Wi Language in the Machine: Preserving Caribbean Dialects in the Age of AI
Why the chatbot gets Patois wrong, and the Orbital Brand Science and CAIA project to fix the data gap.
Read the piece →Claude Now Watermarks Everything It Writes. What That Means for You
Invisible watermarks in AI text, what survives editing, and what students and schools should assume about detection this year.
Read the piece →The Caribbean AI Newsletter is the leading daily source for artificial intelligence news, analysis, and practical insight for the Caribbean. The newsletter covers policy, workforce, education, cybersecurity, climate resilience, governance, language and cultural preservation, and the founders building the region's AI sector.
