evidoria

← Back to browse

Good practice

Standards and Guidelines for the Use of AI in Assessment — Caribbean Examinations Council's Region-Wide AI Policy for CSEC, CAPE and CCSLC

Barbados · Bridgetown · See the Barbados profile

The Caribbean Examinations Council published binding, region-wide standards governing AI use in CSEC, CAPE and CCSLC school-based assessments across its 16 Caribbean territories from 2026, naming approved AI tools, originality checkers and mandatory disclosure rules.

Standards and Guidelines for the Use of AI in Assessment — Caribbean Examinations Council's Region-Wide AI Policy for CSEC, CAPE and CCSLC Standards and Guidelines for the Use of AI in Assessment — Caribbean Examinations Council's Region-Wide AI Policy for CSEC, CAPE and CCSLC

Details

Promoter
Caribbean Examinations Council (CXC)
Period
Published Apr 2025; applies to school-based assessments from the 2026 examination cycle
Keywords
AI governance, academic integrity, assessment policy, generative AI, data protection

Context

The Caribbean Examinations Council (CXC), the statutory body headquartered in Bridgetown, Barbados that sets and administers the CSEC, CAPE and CCSLC secondary examinations across its 16 Caribbean member territories, published its 'Standards and Guidelines for the Use of Artificial Intelligence in CXC's Assessment' (V1.0) in April 2025. The policy was announced at CXC's 4th Ministerial Summit in Trinidad in October 2024.

Objectives

The standards aim to establish common, enforceable rules for academic integrity and data handling ahead of AI's arrival in high-stakes regional School-Based Assessments (SBAs), governing how generative AI may be used by candidates and by CXC itself in marking.

Activities

The policy names specific permitted AI engines (including Copilot, Gemini and ChatGPT) and approved originality-checking tools (including Turnitin, GPTZero and Originality.ai); requires candidates to complete a CXC SBA AI Disclosure Form acknowledging any AI use and confirming final submissions reflect their own interpretation and work; and sets expectations that any AI CXC itself uses to mark candidates' work be trained and retrained on representative data to produce consistent results. CXC Registrar and CEO Dr Wayne Wesley has framed the effort as part of a wider push to infuse AI across qualifications, syllabus, teacher development, assessment and administrative processes.

Conclusions

The standards take effect for candidates sitting CSEC, CAPE and CCSLC school-based assessments from the 2026 examination cycle onward, making this one of the first region-wide, binding AI-assessment governance frameworks to apply across an entire multi-country examination system. As a governance instrument, it does not itself produce learning-outcome data; its contribution is procedural — establishing rules ahead of AI's arrival in assessment.

Implementation
Indicative cost
Low (< €50k) — No budget figures are published in available sources; the initiative is a standards/policy document rather than a funded technology build, so cost is estimated conservatively as low at the CXC secretariat level (tool licensing costs, if any, would fall on individual schools/territories).
Time to results
Medium (1–3 years) — Announced at CXC's 4th Ministerial Summit in October 2024, standards published (V1.0) in April 2025, taking effect for the 2026 examination cycle onward.
Staffing & skills
Caribbean Examinations Council (CXC) secretariat, headquartered in Bridgetown, Barbados, CXC Registrar and CEO Dr Wayne Wesley (public sponsor of the initiative)

Conditions for success

  • Region-wide, binding applicability across all 16 CXC member territories rather than a single school or ministry, avoiding a patchwork of local rules
  • Explicit naming of approved AI engines and originality-checking tools, giving schools and candidates clear, actionable guidance rather than abstract principles
  • Mandatory candidate disclosure form creating a documented audit trail for AI use in assessments

Common failure modes

  • As a policy document, the standards do not themselves report or measure any change in learning outcomes or academic-integrity incidents, so their real-world effectiveness is not yet demonstrated
  • Reliance on named third-party AI-detection tools (e.g., Turnitin, GPTZero) carries some risk given known accuracy limitations of AI-content detectors generally, though this is not addressed in the source material

Where it fits

Governance type
regional statutory examinations body governing multiple sovereign member territories
Scale
regional (16 Caribbean territories)
Income level
mixed (predominantly middle-income Caribbean states)

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful