A complete integration guide for the Coalition for Content Provenance and Authenticity standard. SDK setup, manifest validation, AI detection, and platform policy enforcement.
The Coalition for Content Provenance and Authenticity (C2PA) defines an open standard for content credentials. The standard records what tool created a piece of content, who edited it, and the chain of custody, all as a cryptographic manifest attached to the content itself.
Most user-generated content today does not carry C2PA manifests. As adoption grows (Adobe, OpenAI, Microsoft, BBC, and others are now issuing manifests), platforms need to validate the manifests when present and detect AI-generated content when manifests are absent.
Content with a C2PA manifest. Validate the manifest against the issuer's signature, confirm the manifest has not been tampered with, and surface the trust signal.
Content without a C2PA manifest. Run AI detection across image, video, and audio modalities. Surface the detection verdict.
The combined approach. Every piece of content gets evaluated on both tracks. C2PA tells you the claimed history. AI detection tells you the actual content type. Both contribute to the final trust signal.
deepidv provides SDKs for the four major platform contexts: Web (JavaScript), iOS (Swift), Android (Kotlin), and server-side (Node.js, Python, Go). Each exposes a single evaluate method that runs both tracks in parallel and returns the combined verdict.
Every evaluation returns a structured response with both tracks: c2pa (manifest_present, manifest_valid, issuer, claim_signed_by, claim_chain, tampering_detected) and ai_detection (per-modality verdict with confidence and probable_generator). The combined_verdict synthesizes both tracks.
The combined verdict is one of seven outcomes: authentic_with_c2pa, authentic_without_c2pa, ai_generated_with_c2pa_disclosure, ai_generated_without_disclosure, edited_with_c2pa, edited_without_c2pa, and tampered_or_invalid. Platform policy is configured per content surface — a news organization may apply different policies than a marketplace.
Each platform configures policy per verdict outcome and per content surface. User profile photos can have different policy than news articles, and both can have different policy than marketplace listings. The policy engine runs at the upload boundary or in async post-processing depending on the surface's latency tolerance.
Platforms with existing content libraries can run backfill to evaluate historical uploads. The backfill API accepts batches of up to 1,000 items and returns evaluations within 5 minutes per batch. Backfill pricing is volume-based with enterprise discounting at 100M+ asset thresholds.
The integration aligns the platform with multiple emerging standards: C2PA Content Credentials, EU AI Act Article 50, UK Online Safety Act, Adobe Content Credentials, Coalition for Content Provenance, Content Authenticity Initiative, and Project Origin (BBC, Microsoft, NYT). Documentation aligned to each standard auto-generates per evaluation.
C2PA manifest validation in under 50ms. AI image detection in under 200ms. AI video detection in under 800ms per minute of video. AI audio detection in under 400ms per minute of audio. Combined verdict synthesis in under 10ms. For platforms doing 100M+ uploads per day, deepidv operates dedicated infrastructure with tenant isolation.
What is C2PA? The Coalition for Content Provenance and Authenticity. C2PA defines an open standard for content credentials that travel with content as a cryptographic manifest. The manifest records what tool made the content, who edited it, and the chain of custody.
Do I need to be a C2PA member to validate manifests? No. Manifest validation is open. Any platform can validate manifests against issuer signatures using the open SDK.
What happens to content uploaded before C2PA was widely adopted? Run backfill to evaluate the existing library. AI detection runs regardless of manifest presence. Backfill tags historical content with the same verdict structure as live uploads.
Can I customize the policy per content surface? Yes. Policy is configured per verdict outcome and per content surface. A user profile photo can have different policy than a news article, and both can have different policy than a marketplace listing.
How does this align with EU AI Act content labeling? The EU AI Act Article 50 requires platforms to label AI-generated content. The combined evaluation produces both detection and (when present) C2PA disclosure. Platform policy can be configured to apply Article 50 labeling automatically based on the combined verdict.
What's the cost model? Volume-based per evaluation with enterprise discounting at 1M+ evaluations per month. Backfill pricing is separate and quoted per project.
Go live in minutes. No sandbox required, no hidden fees.
Building vs. buying identity verification infrastructure is one of the most consequential technical decisions a growing company makes. Here is the framework for getting it right.
Evaluating identity verification providers? This comprehensive guide covers every criterion that matters — from technical capabilities to pricing models to vendor stability.
Monolithic KYC bundles force you to pay for checks you do not need. Modular identity verification lets you compose workflows that match your exact requirements — and nothing more.
