Deepfake
AI-generated synthetic media — typically video or audio — that convincingly impersonates a real person doing or saying something they did not.
Deepfakes use generative models to map a target's face or voice onto source footage. The 2025–2026 generation can produce real-time deepfaked video calls and voice cloning from under 30 seconds of audio. Detection has not kept pace with generation; the most reliable defense is content provenance — cryptographic signing of camera output (C2PA standard) and verified-source backtracking. For news consumers, three signals raise the deepfake flag: (1) the only copy circulating is on a single platform, (2) no major outlet has independently confirmed the footage, (3) the alleged source explicitly denies it. See also: misinformation, content provenance.