OSINT & investigations
Investigators collect photos, screenshots and clips from social media, messengers and seized devices. A growing share of it is generated or manipulated. Our API scores every file for AI generation from the pixels, so analysts know what to trust before it goes into a report.
99.5%
of AI images caught with only 1% of real photos flagged, across 19 current generators
96.2%
still caught after WhatsApp, Instagram and re-share compression
0.987
AUROC on AI video from 12 video models
Hidden split of our October 2026 benchmark, never used in training. Method and full results
By the time content reaches an analyst it has been forwarded, screenshotted and recompressed. EXIF data and content credentials are usually stripped, so provenance tools have nothing to read.
Current image and video models produce faces, documents and scenes that trained eyes can no longer reliably tell apart from camera footage.
A single case can hold thousands of media files. Nobody can inspect each one by hand, and a missed fake can undermine an entire line of inquiry.
A probability that an image was generated or synthetically altered, trained on the newest generators and on copies that went through WhatsApp, Instagram and re-sharing.
Frames are sampled across the clip and scored by a dedicated video model, with the number of frames that looked AI-generated, so you can see where in a clip the signal sits.
A thumbnail or a screen grab from footage can be checked on its own with the video model, without the original file.
Where a file still carries a C2PA manifest, it is parsed and its signature verified. A valid AI declaration from the signer is reported as certainty, not a score.
Upload images and clips, or pass a URL, from your collection or case-management pipeline through a REST API.
Set thresholds to your tolerance for false positives: flag high scores automatically, route the grey zone to an analyst, clear the rest.
Store the score, the frame counts and any content credentials alongside the item, so the assessment is documented with the evidence.
Score a clip from a case
curl -X POST https://www.wasitaigenerated.com/api/v1/detect/video \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://example.com/evidence/clip.mp4"}'Images, single frames and whole clips. See the API docs for responses, limits and pricing.
Yes. Police, prosecutors, election bodies and other public agencies use the same API to check images and video in investigations and disinformation monitoring. For sensitive work we offer a data processing agreement, dedicated capacity and self-hosting on your own GPUs and servers, so files can stay in your own environment.
Treat it as an investigative lead, not proof. The score is a probability from a model. Pair it with other checks such as source tracing, reverse search and content credentials, and have an analyst review anything a decision depends on.
Yes. The model is trained on images that went through messenger and social-media recompression. On our published benchmark it still caught 96.2% of AI images after WhatsApp, Instagram and re-share compression, at a 1% false-positive rate.
Uploaded files are deleted after analysis; the API keeps the result, not the file. A data processing agreement and self-hosting on your own GPUs and servers are available for sensitive casework.
Yes. Enterprise customers can run our models on their own GPUs and servers, on-premise or in their own cloud, so files never leave their infrastructure. We also offer dedicated capacity in our cloud. Talk to us about your requirements.
We run paid pilots on a sample of your images and clips, calibrate the threshold to your tolerance for false positives, and report detection and error rates back before you commit. Volume pricing and an SLA are available for production, and the models can run on your own GPUs and servers.
Talk to us about a pilot