Truepic, a San Diego-based image authentication startup used by Reddit to verify identities in AMAs, raises $8M Series A to tackle video deepfakes
Josh Constine / TechCrunch : Tweets: @joshconstine Tweets: Josh Constine / @joshconstine : Deepfakes could start a war, but this cool startup is building ways to detect them via discrepancies in hair, ears, & reflectivity of eyes http://techcrunch.com/...
Context & Ripple Effects
Truepic's pitch in 2018 was unusual among authenticity startups: rather than spotting fakes after upload, it verifies images at the moment of capture using proprietary algorithms — an approach MIT Technology Review profiled alongside UK rival Serelay that same year (capture-time verification). Its only named customer at the Series A stage was Reddit, which used the tool to verify identities for AMAs, giving the startup a live consumer-platform deployment before most buyers had a deepfake budget.
The $8M round reads differently in hindsight: three years later Truepic raised a $26M Series B led by Microsoft's M12, confirming that capture-time provenance — not just detection — became a funded category as synthetic media spread.
First-order effects
- The new capital lets Truepic extend its capture-time verification from still photos into video, directly targeting the deepfake problem its TechCrunch profile frames via artifacts like hair edges, ears, and eye reflectivity.
- Reddit's AMA identity-verification workflow gains a path to cover video content, deepening the platform's reliance on third-party authentication rather than in-house tooling.
Second-order effects
- Rivals must pick a lane: Serelay competes head-on in capture-time verification, while post-hoc detectors like Intel's FakeCatcher — which claims 96% accuracy reading pixel-level 'blood flow' (real-time deepfake detection) — push the opposite thesis that fakes can be caught after the fact.
- Adjacent identity firms such as AuthenticID, which raised $100M for AI-based proofing, blur the line between document fraud checks and media authentication, pressuring pricing across both markets.
Third-order effects
- The market is structurally splitting into two camps — provenance baked in at capture versus forensics applied after upload — and as the Washington Post's survey of detection startups notes, untested accuracy claims make provenance the more defensible long-term anchor.
- If platforms keep outsourcing authenticity (as Reddit did for AMAs), verification becomes purchased infrastructure, consolidating around a few funded vendors rather than being built in-house.
The trend: Media authenticity is becoming a layered trust stack, with capture-time provenance vendors like Truepic competing against post-hoc deepfake detectors for the same platform budgets.