What is Deepfake?

A deepfake is a piece of synthetic media — usually video or audio, sometimes images — in which a real person's face, voice, or likeness has been digitally fabricated or swapped, typically without their consent. The term originally referred to face-swap videos produced by the deep-learning model that spawned the early 2018 reddit communities, and has since expanded to cover any high-fidelity AI-generated impersonation of a real identity. Deepfakes are categorically distinct from AI UGC and brand-owned AI personas: the legal and ethical risk lives in 'impersonating an identifiable real person,' not in 'using AI to generate visual content.' Deepfakes carry meaningful legal exposure under right-of-publicity statutes (every US state recognizes some version), federal trademark/false-endorsement law, the EU AI Act, the UK Online Safety Act, and a growing list of state-level deepfake-specific statutes (Tennessee ELVIS Act, California AB-2655 and AB-2839, etc.). Detection technology (Intel FakeCatcher, Reality Defender, Sensity AI, Microsoft Video Authenticator) and provenance standards (C2PA, SynthID, IPTC AI metadata) are converging to make synthetic-content origin verifiable, but consumer-side awareness is the more durable defense. Brands should never use deepfake-style identity-swap technology in ad creative — even when the underlying tool is commercially available — because the brand-risk asymmetry makes any short-term creative gain catastrophically expensive.