Evidence
Voice clones target the elderly in China and the US alike
| Date | Case | Country | What happened | Source |
|---|---|---|---|---|
| Jan 2024 | Arup deepfake CFO call | China (Hong Kong) | 15 transfers totaling HK$200 million | Source |
| Jun 13, 2025 | Biden New Hampshire robocall | United States | FCC fined the organizer $6 million; a state jury acquitted him of criminal counts | Source |
| Late 2025 | Beijing grandson voice-clone scam | China | An elderly woman paid 80,000 yuan to a cloned grandson; eight suspects detained | Source |
| 2025 | Crypto ATM impersonation fraud | United States | $333 million in deepfake-linked losses, mostly older adults | Source |
Registry
No registry incident turns on impersonation yet
The cases above are documented, but none yet meets the registry bar of a named developer, a failed control and a known finder.
Exposure
Eight of 17 agent types rate very significant or critical on identity and authentication
From the attack surface ratings, the agent types where this risk has the widest path in.
| Agent type | Dimension | Rating |
|---|---|---|
| Enterprise agents | Identity and authentication | Critical |
| Robotaxis and autonomous vehicles | Identity and authentication | Very significant |
| Fleets and swarms | Identity and authentication | Very significant |
| Phones and tablets | Identity and authentication | Very significant |
| Smart home assistants | Identity and authentication | Very significant |
| Coding and research agents | Identity and authentication | Very significant |
| Workflow agents | Identity and authentication | Very significant |
| Commerce agents | Identity and authentication | Very significant |
Controls that bound it
- Identity and Cryptographic Attestation: who or what is acting, proven cryptographically
- Consent and Capture: what always-on devices record, signal and keep
Rules
China labels synthetic media; the US criminalizes intimate deepfakes and AI robocalls
| Rule | Jurisdiction | What it requires | Status |
|---|---|---|---|
| AI-generated content labeling measures | China | Explicit and metadata labels; platform checks | Effective Sep 1, 2025 |
| Deep synthesis provisions | China | Consent for face and voice editing; labeling; real-name users | Effective Jan 10, 2023 |
| Take It Down Act | US | Intimate deepfake crimes; 48-hour removal | Platform duties from May 19, 2026 |
| FCC ruling on AI voices | US | AI voices count as artificial under the TCPA | Feb 8, 2024 |
| EU AI Act Article 50 | EU | Deepfakes must be disclosed | Aug 2, 2026 |
Status colors: green in force, blue voluntary standard or framework, amber adopted and phasing in, gray draft or pending, red withdrawn.
Record
Page Record
- Risk group
- Human Risks
- Datasets
- ai_risk_evidence.csv (4 rows); registry incidents (0 linked); ai_attack_surface.csv
- Key sources
- Arup deepfake CFO call
Biden New Hampshire robocall
Beijing grandson voice-clone scam
Crypto ATM impersonation fraud - Last verified
- October 2026