AI Image Detector
API Key
DocsPricingUse CasesBlog
Get API Key →
Changelog

Every change, in public

Model updates, new endpoints, SDK releases, security disclosures, and behavior changes. Subscribe to the RSS feed if you build on us.

RSS feedEmail updatesView status
  1. Model

    Detection model v4.2.0 deployed

    • •Added Sora-image v2 attribution. Per-model accuracy now 94.6% (up from 92.1%).
    • •Improved adversarial robustness on recompressed images by 3.4 percentage points.
    • •Reduced false-positive rate on heavily edited real photos from 1.8% to 1.4% at the 0.7 threshold.
    • •Region heatmap resolution increased from 32x32 to 64x64 for better inpainting localization.
  2. Feature

    C2PA verification endpoint /v1/c2pa/verify is now generally available

    • •Cryptographic verification of Content Credentials manifests with full signing chain validation.
    • •Returns trust assessment plus full manifest contents.
    • •Free on all tiers, including Free.
  3. SDK

    .NET SDK 1.0 released

    • •Seventh official SDK. Targets .NET 6.0+.
    • •Same automatic retries, request signing, and response parsing as our other SDKs.
    • •Install via dotnet add package AIImageDetector.
  4. Feature

    Webhook delivery hardening

    • •Per-destination rate limiting added (default 100 RPS, configurable on Enterprise).
    • •Exponential backoff with jitter on 5xx responses.
    • •Failed deliveries now retried 5 times over 24 hours (was 3 over 6h).
    • •X-Idempotency-Key header for safe retry handling.
  5. Fix

    EU region routing fix

    • •Resolved upstream provider issue that caused 32-minute latency spike. See status page for details.
    • •Added cross-region failover so customer traffic now automatically routes to nearest healthy region.
  6. Security

    SOC 2 Type II report (FY2025) available

    • •Audited by an AICPA-licensed firm. Period covered: April 2025 to March 2026.
    • •Zero qualifications. Available under NDA at security@aiimagedetectorapi.com.
  7. Model

    Detection model v4.1.0 deployed

    • •Added Flux 1.1 Pro attribution.
    • •Improved DALL-E 4 detection accuracy from 98.9% to 99.5%.
    • •Better handling of cropped and aspect-ratio-shifted AI images.
  8. Feature

    Async batch jobs API

    • •New /v1/detect/async endpoint with webhook callbacks for high-volume batch detection.
    • •Up to 10,000 images per async job (Pro), unlimited on Enterprise.
    • •Job status polling via /v1/jobs/{id}.
  9. Feature

    Pricing page launched with self-serve Pro tier

    • •Pro tier at $49 per month for 50,000 scans is now self-serve via dashboard.
    • •14-day free trial on Pro, no credit card.
    • •Annual billing option with two months free.
  10. Model

    Detection model v4.0 deployed

    • •Migrated from CNN-based architecture to vision-transformer.
    • •Aggregate accuracy improved from 97.8% to 99.1% on internal benchmark.
    • •p50 latency reduced from 142ms to 87ms.
  11. SDK

    Java SDK 1.0 released

    • •Sixth official SDK. Gradle + Maven support.
  12. Fix

    Dashboard cache fix

    • •Stale cache entries caused slow page loads. Cache layer rebuilt, TTL reduced from 60s to 15s.
  13. Deprecation

    Detection model v3.x deprecated

    • •v3.x model will be retired on March 15, 2026.
    • •All v3 endpoints continue to work but return Deprecation warning headers.
    • •Upgrade by setting model_version=4 in request body. v4 is the default for all new keys.
  14. Feature

    Region heatmap endpoint generally available

    • •Pixel-level heatmap output (was beta since Q3 2025).
    • •Returns a 32x32 grid of suspicion scores plus a downloadable PNG visualization.

Want to influence the roadmap?

Enterprise customers get a quarterly review with our product team. Pro customers can request features at feedback@aiimagedetectorapi.com.

See plans