ScamAI raised $2.6M to combat AI-powered scams
scam.ai

~/docs cat document-forgery.md

Document forgery

POST /api/defence/docforge/detect — detect tampering and forgery in document images, with region-level evidence and visual overlays.

Request: file (binary, required) — JPEG, PNG, WEBP, HEIC/HEIF, or TIFF, max 50 MB; optional save and caller-supplied task_id.

curl
curl -X POST "https://api.scam.ai/api/defence/docforge/detect" \
  -H "x-api-key: <YOUR_API_KEY>" \
  -F "file=@/path/to/document.jpg"
response fields (key)
result.verdict           "LIKELY_AUTHENTIC" | "LIKELY_FORGED"
result.confidence        confidence in the verdict, 0.0–1.0
result.cal_prob          calibrated forgery probability, 0.0–1.0
result.forged_area_pct   % of image area flagged as forged
result.n_regions         suspicious regions detected
result.regions[]         per-region boxes: x, y, w, h, area_pct, peak_prob
result.resolution_tier   unreliable | low | acceptable | optimal | high
result.overlay_png_b64   base64 PNG — flagged regions outlined
result.heatmap_png_b64   base64 PNG — forgery heatmap
credits_used             1 per image

Recoverable per-file failures return 200 with result.success: false and an error code: RESOLUTION_TOO_LOW, UNSUPPORTED_TYPE, FILE_TOO_LARGE, or INFERENCE_ERROR. See ID document forgery and insurance claims.