Image
78.1% of 653 AI images from acceptance-v4, scored through the deployed pipeline drew a strong warning. This source-asserted slice is not an overall accuracy figure for the deployed lane.
Benchmarks and limitations
Evidence.
Including the misses.
No current lane supports a broad 95% accuracy claim. Results below are warning-lane evidence, not an authenticity certificate.
Open the free checkerScope before score.
Every result is a warning signal. The evidence and limitations below define what this route can and cannot support.
Sampled video
Through the deployed lane, 3.5% of 200 official-trailer frames from 70 trailers across 6 unseen distributors drew false warnings.
Historical artifact
Earlier per-model aggregates (V7, V10, V12) are superseded. They do not describe the deployed routed lane and must not be quoted as current.
Lanes that are not in the public product
Image and sampled video are the public product. Text and audio are not marketed as supported public lanes.
Do not infer more than the lane measured.
- No overall accuracy claim is supported by these bounded slices.
- Source attribution is not cryptographic provenance.
- Sampled video aggregates frames and is not temporal deepfake analysis; moments the sampler never lands on are never scored.
- Text and audio are not public product lanes.