Definition
Fake images, video, audio, or text auto-created by generative AI that can deceive people or damage a business.
At a glance
- Spans deepfake video, cloned voices, fake photos, and convincing text, all now cheap and fast.
- A direct fraud threat: in 2024, engineering firm Arup wired ~$25M after a video call with deepfaked executives.
- People rarely catch it, spotting high-quality deepfakes only ~24.5% of the time.
- U.S. AI-fraud losses are projected to climb from $12.3B in 2023 to $40B by 2027.
Why it is different now
Misinformation is false info that spreads; disinformation is the deliberate version[1]. The tool changed it. Generative AI now makes a realistic fake video, clones a voice from a short clip, or writes a flawless scam email in seconds, for almost nothing, so any fraudster can impersonate your CFO or fabricate news about you[4]. Deepfake files jumped from ~500,000 in 2023 toward 8 million in 2025[2].
How it hits a business
The costliest form is impersonation: scammers use deepfaked audio or video of a leader to order a wire or data, often live on a call[3]. Arup lost ~$25M; an energy firm sent ~$243,000 to a voice clone of its parent CEO[2]. AI also drives fake reviews, cloned sites, and tailored phishing.
How to protect yourself
Set a code word for urgent money requests. Treat urgency and secrecy as red flags. Detection software is unreliable, so use it only as backup[5].
Bottom line
You cannot win by looking harder, so build a verification habit and slow urgent requests until they prove out.