AI use and disclosure

Wits World uses AI assistance for code, editing, review, and documentation. Contributors remain responsible for the work they accept. The supplied COMS3011A Project Brief, section 1.4, requires attribution of the tool, model, and purpose and directs contributors to the separate course AI policy for the full requirements.

Recorded uses

The repository README records code generation, inline editing, and code review separately. Its declarations include the following tools and models:

Tool	Recorded models
ChatGPT-Web	GPT-5.6 Thinking; GPT-5.6 Sol.
Codex	GPT-5; GPT-5.6 Sol; GPT-6.
Codex-CLI	gpt-5.6-sol medium; GPT-5; GPT-6; gpt-6-sol; gpt-6-luna.

This list summarises project-level disclosure. Individual changes identify which tool and model contributed and the purpose of that assistance.

Documentation records also identify AI-assisted planning, prose generation, source comparison, and editing. For example, the Sprint 3 guides attribute drafting and source verification to Codex-CLI[gpt-6-sol] and Codex-CLI[gpt-6-luna]. The README distinguishes that work from application code generation. The September documentation audit attributes navigation, test-code changes, rubric/source comparison, and evidence review to Codex-CLI[GPT-6].

Attribution

The README summarises repository-wide uses. Assisted code commits name every materially contributing tool and model in an Assisted-by trailer, while the PR's AI section describes the assistance with that change. Submission-facing documents include a declaration naming the tool, model, and purpose of their generation, planning, review, or editing.

For example, this documentation revision can be described as:

Codex[GPT-6] assisted with rubric comparison, repository source review, rewriting the Process guides, and documentation validation.

The matching commit trailer is Assisted-by: Codex[GPT-6]. Contributors use the actual tool and model rather than copying an example or the full project list. Unknown model details stay unknown until supplied.

Verification and responsibility

Contributors inspect the diff, understand the resulting behaviour, check facts against source material, and run checks appropriate to the change. Human review and the quality gates apply to AI-assisted work. A proposed review finding needs confirmation against the code or a reproducible scenario before it becomes a reported defect.

Generated code and documentation can contain incorrect assumptions, stale contracts, or unsupported claims. Source review and tests help identify those errors; attribution does not establish correctness. Contributors also check privacy, security, licensing, accessibility, and operational consequences.

AI-generated text cannot supply missing participant observations, stakeholder approval, meeting attendance, or test results. Reconstructed sprint summaries retain their source limits, and verification records identify the revision, environment, procedure, and actual outcome. Credentials, personal participant data, and private location histories stay out of public declarations.

AI declaration: Earlier versions used Codex[GPT-5], ChatGPT-Web[GPT-5.6 Sol], and Codex-CLI[GPT-6] for generation, editing, and review. This revision used Codex[GPT-6] for rubric comparison, source review, rewriting, and documentation validation.