San Francisco Home Invasion: $11M Cryptocurrency Stolen at Gunpoint
BlockedPhysical coercion was applied — the custody structure did not protect against forced transfer.
In November 2025, an armed robber entered a residential home in San Francisco by posing as a delivery worker. The attacker subdued the homeowner by tying them up and used physical coercion to force the transfer of approximately $11 million worth of cryptocurrency. The case was reported by the San Francisco Chronicle.
This incident illustrates a vulnerability specific to self-custody models: the human operator remains a single point of access risk. Unlike institutional custodians with security protocols, surveillance, and insurance frameworks, self-hosted cryptocurrency is defended only by the owner's physical security and operational discipline. The attacker's method—social engineering entry as a delivery worker—exploited ordinary residential trust patterns.
The outcome and recovery status of the stolen assets remain unknown as of the archival date. No public reports have documented whether law enforcement recovered any portion of the funds, whether the attacker was apprehended, or whether blockchain analysis traced the transferred cryptocurrency to identifiable exchanges or wallets.
This case has implications for estate and security planning: holders of substantial self-custodied assets face not only the traditional custody risks (passphrase loss, device failure, documentation gaps) but also physical security threats that can force immediate asset transfer under duress. The vulnerability window extends beyond the owner's lifetime—an attacker with knowledge of holdings could target family members or heirs.
| Stress condition | Coercion |
| Custody system | Hardware wallet (single key) |
| Outcome | Blocked |
| Documentation | Present and interpretable |
| Year observed | 2025 |
| Country | United States |
What custody structure can and cannot protect against coercion
The relevant structural question is not whether a custody setup can prevent coercion — it typically cannot — but whether it can limit what an attacker can obtain through coercion. A setup where the holder has sole knowledge of all credentials, with no geographic distribution and no multisig threshold, gives an attacker everything they need by controlling one person. A setup where credentials are geographically distributed, where multisig requires coordination with parties in other locations, or where a passphrase-protected decoy wallet exists, limits what any single physical attack can yield.
Observed cases in this archive range from violent home invasions and kidnappings to subtler forms of coercion: legal threats, family pressure, business disputes that escalated. The outcomes depend on whether structural protections existed and whether they held under pressure. Setups with no geographic distribution or threshold requirements produced the worst outcomes.
The legal dimension adds complexity: transactions executed under coercion are technically valid. The blockchain cannot distinguish voluntary from involuntary signatures. Recovery after a coerced transfer depends entirely on legal processes — identifying the attacker, prosecuting, and attempting asset recovery — which is slow, expensive, and uncertain.
The most effective structural protection against coercion is geographic key distribution combined with a signing threshold that cannot be met from one location. An attacker who controls one person in one place cannot force a transaction that requires coordination with key holders in other jurisdictions. This protection requires accepting coordination overhead during normal use.