U.S. authorities say they have disrupted cloud infrastructure supporting a North Korean cybercrime operation that used deceptive video calls, stolen accounts and artificial deepfakes to target people in cryptocurrency and software development. The campaign, called FaceStealer by the FBI, is a reminder that an apparently familiar face on a call is no longer sufficient evidence of a familiar person.

In a November 12, 2026 announcement, the Justice Department said the FaceStealer scheme had been linked to more than $400 million in cryptocurrency theft over the previous 18 months. The FBI's dedicated advisory says agents identified and seized more than 900 webcam videos of victims working in crypto and Web3 industries around the world.

The announcement also described separate court-authorized actions involving more than $1.2 billion worth of cryptocurrency stolen from Binance. Those asset actions should not be added to the FaceStealer tally as if they were the same set of victims or the same single loss figure.

How trust was turned into an attack surface

According to the FBI, attackers first compromised people and secretly collected video material, then hijacked communication accounts to contact their victims' professional networks. A new target might encounter a video-conference invitation apparently sent by a colleague and see a convincing representation of a familiar contact.

Authorities say the operation combined prerecorded material with deepfake content. Once a target trusted the interaction, the attackers sought to induce risky action on the target's device. The important distinction is that this was not merely a fake image shared on social media: it was impersonation embedded into a live-looking business workflow, backed by compromised accounts and real contact lists.

The FBI attributes FaceStealer to North Korean-linked actors known by several names in private threat reporting, including Sapphire Sleet and BlueNoroff. These attributions and financial totals are law-enforcement findings, not independent audits of every affected transaction. The agencies have not claimed that all stolen funds have been recovered or that the broader network of perpetrators has been eliminated.

Why open-source projects are in the warning

Beyond cryptocurrency theft, the FBI says the same techniques increasingly target maintainers of widely used open-source software. That expands the potential consequences: access to a maintainer's repository or publishing account could become a route into software relied on by companies that never interacted directly with the fraudsters.

The risks are distinct from the problems addressed by conventional model safety testing. A project can have secure source code and still suffer a supply-chain compromise if a trusted contributor's identity is stolen. Likewise, multifactor authentication and security awareness are useful controls, but neither guarantees protection when a genuine session or a legitimate human operator is manipulated.

What organizations can verify

Teams can reduce dependence on a single video call as proof of identity. Sensitive changes—such as repository access, payment approvals, credential resets or installing unfamiliar tools—should require an independent confirmation channel and narrowly scoped permissions. Administrators should monitor unusual account activity, use hardware-backed authentication where suitable, and treat unexpected meeting requests as potentially untrusted even when the sender appears known.

These are risk-reduction measures, not proof that a particular organization has been targeted. The FBI's advisory specifically warns that compromised communication accounts can make fraudulent invitations appear to originate from trusted contacts. Independent confirmation helps break that trust chain before a sensitive action takes place.

The strongest takeaway from the disruption is that deepfakes are most dangerous when paired with ordinary account compromise and organizational trust. Disrupting one cloud operation is meaningful, but sustained protection depends on controls that do not assume the face, voice or account on a screen is enough.

Sources: U.S. Department of Justice, November 12, 2026; FBI FaceStealer advisory.