Verify before it spreads.
Vuka Peace checks claims against evidence, explains what's misleading, and helps human mediators respond — calmly, factually, before rumour becomes tension. It does not monitor communities, score risk by group, or notify authorities.
One path: detect, verify, explain, respond, mediate.
No path in Vuka Peace scores a community or routes to law enforcement. Every claim ends with either a sourced answer or a human conversation.
Detect claims
A claim is submitted by a community member or spotted in public sources — never scraped from private accounts or messages.
Verify evidence
Independent sources are cross-checked: news reporting, official statements, fact-checking organisations, and prior verified claims.
Explain misinformation
If a claim is false or misleading, the platform explains why in plain language, with the evidence shown — not just a verdict.
Prepare neutral responses
AI drafts a factual, non-inflammatory message a mediator could share — never blame language, never collective accusation.
Human mediation
A trained human mediator reviews, edits and decides whether and how to share the response with the community.
The Truth & Verification Centre.
Every claim below is fictional demonstration data, generated for this prototype. Nothing here is a real report, a real region's activity, or a real person's claim.
DEMO DATA — this feed is not connected to any real claim source. It's built so an authorised fact-checking or community-reporting connector can plug in later.
Different models check each other's work.
No single model decides whether a claim is true. Three independent model families each produce an assessment; a human fact-checker makes the final call.
Deep reasoning & drafting
Used for the hardest part of the job: reading conflicting evidence, explaining nuance, and drafting mediator briefings in plain, calm language.
- Claude Opus 4.8 — ambiguous claims, conflicting evidence, sensitive-context analysis before it reaches a human reviewer.
- Claude Sonnet 5 — everyday claim summarisation, evidence write-ups, mediator briefing drafts, community-facing explanations.
Classification & second opinion
Used to independently triage incoming claims and cross-check Claude's read before anything reaches a human — reducing single-model blind spots.
- GPT-5.6 — a second, independent assessment on claims heading for human review; flags where models disagree.
- GPT Image — generates non-documentary explainer graphics (e.g. "how a rumour spreads") — never used to depict real incidents.
Multimodal evidence checks
Used when a claim includes an image or document — checking for reused photos, mismatched context, or manipulated content.
- Gemini 3.1 Pro — multimodal cross-checking of attached images/documents against known context.
- Gemini 3.5 Flash — fast, low-cost triage of high-volume routine claims before deeper review.
Claim: "An incident reportedly occurred in a fictional demo region."
→ Models agree: sent to human fact-checker with "insufficient evidence" status. No automated conclusion is published.
Understanding xenophobia and misinformation in South Africa.
Rumour and misinformation have repeatedly played a role in South African xenophobic violence. This section is educational and historical — it is not live monitoring data.
Nationwide xenophobic violence
Attacks against foreign nationals across multiple provinces led to deaths and mass displacement, becoming a reference point for later prevention efforts.
Renewed attacks
A further wave of violence targeting foreign-owned businesses and residents was documented in several urban areas.
Johannesburg and Pretoria unrest
Looting and attacks on foreign-owned shops were again reported, prompting renewed civil-society and government response.
Anti-migrant campaigns
Organised anti-migrant rhetoric and operations against foreign nationals were documented by human-rights organisations.
Renewed reporting
Human Rights Watch and Reuters both reported new waves of violence and displacement in 2026 — the direct motivation for this prototype.
South Africa: New Waves of Xenophobic Attacks
African migrants flee into the mountains as xenophobic violence surges
"They Have Robbed Me of My Life": Xenophobic Violence Against Non-Nationals in South Africa
World Report 2025 — South Africa
Tell a human mediator what you're seeing.
Reports go to a human mediator, not an automated risk score. AI may help the mediator summarise and organise what you've written — it never decides whether a group is dangerous.
Spear becomes wing.
The shaft of a spear opens into two wings in flight — vigilance becoming peace. Built as plain geometry, not a generated image, so it stays crisp at any size.
Primary — dark bg
Dark — light bg
White — flat colour bg
Favicon (bolder strokes)
Primary lockup — icon + wordmark
What this platform will never do.
These aren't aspirations — they're hard boundaries on the architecture itself.
Production deployment would require formal legal and privacy review, including assessment against South Africa's Protection of Personal Information Act (POPIA), before handling any real reports or claims.
Who's behind this.
Vuka Peace is a product of Cresent Waters (Pty) Ltd, a private company registered in South Africa.
Cresent Waters
Private Company
Registration No. 2026/659655/07
Krugersdorp, Gauteng
08 Wiege Street
Mindalore, Krugersdorp
Gauteng, 1730
South Africa