CopySight, a startup that checks AI-generated pictures and footage for resemblance to existing intellectual property, has closed a $3 million seed round led by Mucker Capital, with Taisu VC, Flint Capital and Yellow Rocks! also investing. It launched CopyScore V2 alongside the raise, adding video scanning and finer-grained results.
The tool looks for faces, logos, memes and even small design details, then gives each flagged element a probability of matching protected work, along with the owner. In TheWrap's test, a Gemini image of a generic space knight was flagged at 88% for the Mandalorian and 85% for a lightsaber, both credited to Disney. Individual creators use a web version; studios get custom builds.
CEO Artem Petrov argues that AI output is legally ownerless, and wants his score to become the industry standard: something insurers accept before indemnifying a production, and evidence in court. Adviser and investor Doug Shapiro, formerly of Turner, says getting insurers on board early matters because they will not certify many rival tools. Freelance AI artist David Berman runs every stage of his commercial work through it, and says it has caught details as small as a recognisable amplifier in a background. His clients, he says, want proof that the characters he delivers are his own.
Petrov says the service has run about 100,000 checks since January, that more than half of the big studios are testing it and that roughly 350 smaller studios and creators use it, though he named no clients. The pitch comes as Netflix reports generative AI on around 300 titles and Amazon plans AI-made animated series.
Its pitch deliberately targets outputs rather than training data, the side of the AI copyright fight that has drawn less attention so far. Next, the company is building a log of the human steps behind each AI image or video, meant to help studios claim copyright and to credit and pay the creators involved.
Sources · 1