Problem Statement
Since Aug 2026, NPC says AI face-swaps are personal-data processing — posting them without lawful basis can violate Sec. 25 DPA. PNP hunts sharers; students share “funny” edits of classmates and officials that become cyberlibel cases. Victims don’t know how to demand takedown or file with NPC.
Proposed Solution
Upload or paste link: DeepFakeBuster scores manipulation (face-swap artifacts, lip-sync mismatch, metadata), explains in Taglish (“Mukhang AI-swapped, 87% — mukha ni X, hindi niya ito ginawa”), and drafts a platform takedown + NPC complaint with evidence checklist. Schools get a one-page “share = liable” poster. Never identifies unknown people — only verifies user-provided claims.
Target Market
- Primary users: Students, SK, teachers, LGU PIOs, junior reporters.
- B2G: PNP-ACG triage, school guidance offices, COMELEC vs fake endorsements.
How AI Is Used
- Manipulation detector: Ensemble on face artifacts + audio-visual sync; essential because eyes can’t catch modern swaps.
- Explainer + drafter: LLM writes NPC-ready Tagalog complaints citing the Aug 11 2026 notice.
PLDT / Smart / ePLDT / Rokid Angle
- Smart: Feeds confirmed fakes into network scam/deepfake blocking; links to huliscam.
- ePLDT: In-country evidence vault for NPC filings.
- PLDT Enterprise: Brand-protection upsell to banks facing CEO deepfakes.
Demo Scenario (3 minutes for judges)
- Upload mock swapped video (consenting actor): score 87% manipulated with heatmap.
- Taglish explainer + NPC Sec. 25 citation appears.
- Takedown + NPC draft generates in one click.
- Show real video for contrast: “No manipulation found.”
- Close: “Mukha mo, karapatan mo.”
30-Hour Build Plan
| Hours | Task | Role |
|---|---|---|
| 0–3 | Gather open deepfake samples + NPC notice | All |
| 3–12 | Detector + thresholds | AI lead |
| 3–10 | Upload + report flow | Backend |
| 10–20 | Explainer + draft UI | Frontend |
| 20–26 | Red-team with tricky edits | All |
| 26–30 | Pitch, disclosure | PM / Pitch |
Suggested stack: Next.js, deepfake detector, LLM explainer, NPC draft generator. Public data / APIs: FaceForensics++, NPC Aug 2026 notice (public), CICC advisories.
Risks & Privacy Notes
- Face privacy: Processes only user-uploaded media with consent; auto-deletes in 7 days. Demo uses consenting actors/synthetic faces only — no real person’s likeness into AI tools without consent, per NPC rule.
- No doxxing: Never names unknown persons; false-data warnings included.
- IP: Open models with licenses; team prompts owned.
AI Use Disclosure (draft)
Uses deepfake detection plus LLM explanations/drafts. Coding assistants helped with UI. All demo verdicts human-verified; NPC citations checked against the actual notice.
Rubric Scoring
| Criterion | Weight | Score (1–5) | Why |
|---|---|---|---|
| Innovation & creativity | 25% | 5 | NPC-aligned takedown drafter is first locally |
| Problem–solution fit | 25% | 5 | PNP cases + school shares exploding |
| AI application & value | 20% | 4 | Detection + legal drafting need AI |
| Feasibility | 15% | 4 | Open detectors demoable |
| Impact & scalability | 15% | 4 | Nationwide via Smart + schools |
| Weighted total | 4.50 |