Cases › security breach › LR-2026-0905
Coinbase AI scam defenses block deepfake Musk livestream giveaway
Also listed as: AI agent incidents
- Who did it
- Outside attacker
- Who was hit
- more than 500,000 livestream viewers; hundreds of Coinbase customers · United States
- AI product
- Coinbase automated scam defenses
- AI company and model
- Coinbase
- Money lost or records
- US$0
- How it was found
- Automated monitoring
- Reported
- 18 Nov 2026
- Severity
- Moderate
- Sector
- Finance and Insurance
- Coinbase's AI anti-fraud defenses screened customer account activity in production to stop fraudulent transfers, while scammers ran a deepfake Musk livestream promoting a fake Tesla crypto giveaway. The sources do not name the model or the tools it held.
- The AI defenses did not fail; scammers used a deepfake video of Elon Musk to promote a fake Tesla giveaway, and hundreds of customers tried to send millions to them. The boundary crossed was customer funds heading to attacker wallets.
- The livestream reached more than 500,000 viewers, and over roughly two hours hundreds of Coinbase customers attempted transfers of millions of dollars. $4.7 million was stopped from leaving customer accounts.
- Coinbase's automated defenses caught roughly 91% of the attempts without any employee stepping in.
Source: Cryptopolitan · 1 more
What was done
$4.7 million was blocked from reaching the attackers and kept in customer accounts. Neither source states notifications to customers or any control changed afterwards.
Risks Fraud and impersonation (32 incidents) · AI used to run a cyber attack (137 incidents) · Manipulation and misinformation (43 incidents)
No source states: tools it had, company using the AI.
Evidence
- Direct loss $0
Sources (2)
- Coinbase Says AI Stopped $4.7M in Deepfake Musk ‘Tesla Giveaway’ Scam - Cryptopolitan · Cryptopolitan · 2026-11-18
- Scammers Used Musk Deepfake Video to Defraud; Coinbase AI System Blocked $4.7 Million · PANews · 2026-11-18
Show every field, with its source and grade
The AI system
Developer
Coinbase
C one secondary source states it
Software around the model
Coinbase automated scam defenses
C one secondary source states it
What it was allowed to do
Monitoring and blocking Coinbase users' outbound crypto transfers to scammer wallets
D inferred by us from the described behaviour, reason shown
How much it acted alone
it acted alone, unwatched
C one secondary source states it
Stage of use
in live use
C one secondary source states it
Whose system
company that built the AI
D inferred by us from the described behaviour, reason shown
What it was for
Detect and block scam-driven crypto transfers by Coinbase customers
C one secondary source states it
What happened
Earliest source we hold
2026-11-18
D inferred by us from the described behaviour, reason shown
Who caught it
automated monitoring
C one secondary source states it
Near miss
yes, no harm happened
D inferred by us from the described behaviour, reason shown
Who was responsible
an outside party
D inferred by us from the described behaviour, reason shown
Who caused it
an outside attacker
C one secondary source states it
What went wrong
security breach
D inferred by us from the described behaviour, reason shown
AI agent
no AI agent acted
D inferred by us from the described behaviour, reason shown
Impact
Direct loss
$0
D inferred by us from the described behaviour, reason shown
Severity
moderate
D inferred by us from the described behaviour, reason shown
Kind of harm
money
D inferred by us from the described behaviour, reason shown
Who was affected
more than 500,000 livestream viewers; hundreds of Coinbase customers
C one secondary source states it
Why this severity
More than 500,000 livestream viewers were reached and hundreds of Coinbase customers attempted transfers in a live production scam attempt.
D inferred by us from the described behaviour, reason shown
Context
Industry
Finance and Insurance
D inferred by us from the described behaviour, reason shown
Country
United States
D inferred by us from the described behaviour, reason shown
LR-2026-0905 · version 2 · published 2026-11-18 after automatic checks; not yet checked field by field by a person