# CopySight raises $3 million to score IP risk before AI video ships > Source: > Published: 2026-07-29 16:13:36+00:00 Artem Petrov and Konstantin Orlov have raised a $3 million seed round for [CopySight](https://copysight.ai/), betting that generative video will require an automated clearance checkpoint before studios and brands put AI-made footage into commercial production. [Mucker Capital](https://mucker.com/) led the round, with [Taisu Ventures](https://www.taisu.io/), [Flint Capital](https://flintcap.com/?region=united-states) and [Yellow Rocks](https://yellowrocks.vc/) participating, CopySight said Wednesday in a [press release](https://www.prnewswire.com/news-releases/copysight-raises-3-million-to-build-the-intellectual-property-clearance-layer-for-generative-ai-302837194.html). CopySight also launched CopyScore V2, which extends its similarity and risk-scoring system from AI-generated images into video. The financing backs a founder whose career has run through the creative production systems CopySight wants to police. Petrov, CopySight's CEO, previously worked as a creative director at Snap, where a [speaker biography](https://www.brxnd.ai/people/artem-petrov) credits him with work on the creator launch of Spectacles, and later built mixed-reality products at Meta. He studied fine and applied arts in Ukraine and visual communication in Israel before moving into advertising technology and founding companies. That background shapes CopySight's pitch. Petrov is approaching copyright risk as a production problem: creative teams need a machine-readable warning before an asset reaches lawyers, distributors or an audience. Orlov, CopySight's CTO, is building the technical system behind that checkpoint. ### From one image to thousands of frames CopyScore evaluates generated content across five categories: trademarks, characters, brand and iconic designs, celebrity likenesses, and art or visual styles. CopySight says the video product examines footage frame by frame, combines visual matches with prompts and model information, and can flag a logo or recognizable face that appears for only a fraction of a scene. Those transient matches are a practical problem for generative video. A reviewer can inspect one image directly. A commercial video may contain thousands of frames, with different generated objects, backgrounds and faces appearing as the model attempts to preserve continuity. CopySight is selling automation for that review load. The existing [CopySight API documentation](https://copysight.ai/docs) shows how the image product returns detected intellectual property, rights-holder information, bounding boxes and a similarity score between zero and one. CopySight says CopyScore V2 adds multi-pass analysis for ambiguous video segments, adjustable sensitivity thresholds and an immutable record containing prompts, model versions and generation settings. CopySight's public API documentation, viewed on July 29th, still described its verification endpoint as an image-analysis service and said video support was in beta. The documentation may lag the product release, but it shows the work still required to expose CopyScore V2 as a mature integration surface for enterprise production systems. Petrov described the desired output as "objective proof that what they just made is safe to ship." CopySight's legal terms draw a narrower boundary around that promise. Under CopySight's [terms of service](https://app.copysight.ai/legal/terms), the product detects similarities against limited data sources and does not perform legal analysis. CopySight says its reference libraries are not exhaustive, disclaims guarantees that results will be accurate or error-free, and leaves legal clearance with the user. The score is therefore decision support for lawyers and production teams rather than a legal opinion that an asset is non-infringing. That distinction will determine whether CopySight becomes infrastructure or another screening tool. Studios can use a similarity score to prioritize reviews and document their process. They cannot treat a low score as immunity from a rights-holder claim, especially when infringement analysis can involve context, licensing, fair use and jurisdiction-specific law. ### Petrov is selling documentation alongside detection CopySight is pairing similarity detection with a chain-of-creation log that records how an asset was produced and reviewed. That expands the product beyond finding familiar logos or characters. Petrov wants CopySight to document human decisions throughout an AI-assisted workflow, creating evidence that can support ownership and compliance reviews. The approach tracks the U.S. Copyright Office's position on AI-assisted work. In its [January 2025 copyrightability report](https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf?v=20260606082851), the Copyright Office said human-authored expression can receive copyright protection even when a work contains AI-generated material. Purely AI-generated material is not protected, and sufficient human contribution must be evaluated case by case. Prompts alone generally do not establish the necessary control, according to the report. A production log cannot decide copyrightability by itself. It can preserve the prompts, edits, selections and arrangements that a creator may later need to show. CopySight says its records supported Raksha World's successful registration of AI-assisted creative content with the Copyright Office, though the release does not identify the registration number or specify which elements received protection. CopySight says it has processed more than 87,000 copyright and intellectual-property checks since January 2026 and reports a 25x increase in usage over that period. Those figures are self-reported, and the release does not define whether the multiplier compares monthly checks, weekly checks or another usage measure. CopySight also names AGBO, law firm ArentFox Schiff and AI image platform OpenArt as customers, partners or platform users. That grouping spans several types of commercial relationship, but it points to the three constituencies Petrov needs in the same workflow: content producers, legal reviewers and generation platforms. ### The round funds a place inside the production pipeline CopySight says the seed capital will fund deeper integrations into enterprise video production. The business currently stretches from self-service subscriptions to custom enterprise contracts. CopySight's website lists a creator plan at $16 per month when billed annually, a business plan at $160 per month and custom pricing for higher-volume API access, audit logs and private IP libraries. The enterprise push puts CopySight into a market that already includes several forms of automated rights detection. [Higgsfield's Similarity Score](https://higgsfield.ai/apps/similarity-score) screens generated media for possible IP overlap before publication. [Pex](https://pex.com/products/attribution-engine/) identifies copyrighted audio and video for platforms and rights holders, while [YouTube Content ID](https://support.google.com/youtube/answer/2797370) matches uploads against files submitted by copyright owners. CopySight is trying to establish itself earlier in the process, before distribution, and to connect the detection result with evidence about how an asset was made. That positioning gives Petrov a route into studio compliance budgets rather than competing solely as a creator utility. The $3 million round gives CopySight enough capital to pursue those integrations, enlarge its reference data and prove that risk scores can survive scrutiny from enterprise legal teams. Petrov's operating bet is that generative video adoption will depend on clearance systems growing alongside generation models. CopySight now has to show that its scores are accurate enough, documented enough and easy enough to integrate that producers will place them directly in the path between a model's output and a commercial release.