A Plain Memo
Stop trying to spot the fake — switch channels instead
A finance employee in Hong Kong joined a video call with his CFO and five colleagues he recognized — and wired twenty-five million dollars to strangers. None of the people on the call were real.
AI-generated photos, voices, and video now pass the perceptual tests your eyes and ears spent a lifetime running. The honest catch isn't sharper looking — it's structural. Here are the four moves that actually work in 2026, and the one thing worth doing this week.
ℹ️ AI voice disclosure: This episode uses an AI-generated voice (ElevenLabs). The host is a pseudonymous byline persona (Reid Conlin).
In this memo
Sources
- FBI 2025 Internet Crime Report (IC3)
- University of Florida deepfake-detection study, *Cognitive Research*, Jan 2026
- iProov "Deepfake Detection" consumer study, Feb 2025
- Somoray et al., systematic review of 56 deepfake-detection studies, *Human Behavior and Emerging Technologies*, 2025
- NewsGuard, "Leading AI Image Detection Tools Mislead Online Users," May 2026
- Postiglione, Gortner, Subrahmanian (Northwestern), human–AI ensemble deepfake detection, arXiv 2026
- McAfee Labs, "The Artificial Imposter" (3-second voice clone benchmark), 2023
- C2PA Content Credentials
- Arup $25M deepfake fraud, Fortune, May 2024
- FCC enforcement action on the 2024 Biden NH robocall