Patrick Steven Yaroch was a supervisory agent at the FBI. Between late 2024 and early 2025, according to court filings, he used his access to systems tied to a criminal case to move roughly $1 million in digital assets out of wallets belonging to an adversarial nation — assets his own agency was supposed to be safeguarding as evidence. He pleaded guilty this month. He has already forfeited about $925,000 back to government-controlled wallets.
The irony writes itself: an insider with legitimate access to seized crypto decided to treat it as a withdrawal account, and got caught by the same kind of evidence trail his job existed to build.
What the filings describe
According to the case record, Yaroch made ten unauthorized transfers out of the wallets over several months, routing funds through the Sui-based lending protocol Suilend and the Kraken exchange. Investigators later seized devices, seed phrases, and a Trezor hardware wallet from his Virginia residence. In May 2026, court filings say, he queried ChatGPT for investment strategies for the funds — a detail that ended up in the record alongside the transfer history.
He was placed on administrative leave, then terminated, then arrested — all within the same week in August 2026. He has pleaded guilty to theft of roughly $1 million and forfeited $925,000 of it.
This isn't the first time
Insider theft of case-custody crypto by federal law enforcement has precedent, and not a flattering one. In 2015, DEA agent Carl M. Force diverted about $700,000 in Bitcoin during the Silk Road investigation and was sentenced to 6.5 years. That same year, Secret Service agent Shaun W. Bridges stole roughly $350,000 in Bitcoin from the same case and got 6 years. Yaroch's case is the same failure mode a decade later, at a larger dollar figure, with a different agency and a different chain.
The pattern across all three: someone with authorized, in-scope access to a wallet decided the custody controls around that access were optional. That's not a crypto problem. It's an access-control problem that crypto happens to make traceable.
The part that actually matters for anyone holding keys
Here's the detail worth sitting with: a federal agent, with insider knowledge of how investigations get built, still generated a reconstructible trail — wallet-to-exchange hops, wallet-to-protocol hops, timestamps, an AI chat query about what to do with the money. None of that required special forensic magic. It's the same onchain evidence trail that makes any wallet's activity checkable by anyone who bothers to look.
That's the uncomfortable lesson for builders wiring up agent wallets, too: the weakest link in a custody model is rarely the cryptography. It's whoever holds the keys and whether their access is scoped, logged, and limited. "I have legitimate access" is not a security model — for a federal agent or for an autonomous agent with a hot wallet. Spend limits, scoped permissions, and an audit trail aren't paperwork; they're the difference between a contained incident and a $1 million one.
What to watch
Sentencing hasn't been reported yet, and it'll be the next real signal — the Force and Bridges cases both landed multi-year federal terms. Expect renewed scrutiny of how agencies control access to seized-crypto wallets internally, especially as more law enforcement holdings sit in self-custody rather than with exchanges. The forensic playbook here — following wallet-to-protocol-to-exchange flows — is the same one that makes onchain activity harder to hide than cash ever was. That cuts both ways, and this week it cut against the person who was supposed to be the enforcer.
Sources
- ▸[Cointelegraph — Former FBI supervisor admits guilt in $1M crypto theft](https://cointelegraph.com/news/former-fbi-supervisor-charged-1m-crypto-theft?utm_source=rss_feed&utm_medium=rss&utm_campaign=rss_partner_inbound)