In July 2024, a security awareness company did what most companies would consider comprehensive vetting when it hired a software engineer for an internal AI team. KnowBe4's HR team ran four video interviews, completed background checks, and verified references. Then the company shipped him a laptop. Within minutes of the Mac arriving, he began loading malware onto it.

The applicant had used a stolen U.S. identity and an AI-enhanced stock photo. He was a North Korean operative. KnowBe4 caught him because its tooling worked. About 25 minutes passed between the SOC spotting the malware and containing the device, and the company said no data was lost, compromised, or exfiltrated.

Many companies wouldn't have caught it that quickly, and some wouldn't have caught it at all. Two years later, the scheme behind that hire is bigger and better funded, and its operatives are harder to spot.

A fraudulent hire is an intrusion

It's tempting to file North Korea's IT worker scheme under payroll fraud or sanctions evasion. That framing undersells the threat. From a security point of view, a fraudulent hire is an adversary who walks in the front door with valid credentials, a managed device, and a legitimate reason to touch your code. No alert fires.

The laptop farm is the operative's staging server, and a U.S. resident is the proxy.

Recent enforcement shows how large this has become. On October 13, 2026, the Justice Department announced its latest crackdown. Between September 21 and 25, FBI agents searched more than 40 Texas homes believed to be running "laptop farms" for North Korean IT workers and seized about 187 laptops and other devices. An Ohio indictment charges 14 people, and the scheme hit more than 70 U.S. employers. "Each laptop seized potentially represents one company that has hired a DPRK IT worker," said R. Joseph Rothrock, special agent in charge of the FBI's Dallas field office. In a companion public service announcement, the FBI warned that North Korea is infiltrating Western companies with "thousands" of state-controlled remote IT workers, posing a significant national security and intellectual property threat.

Two weeks later, on October 28, the Treasury Department's Office of Foreign Assets Control sanctioned ten individuals and one entity in North Korea and Russia, including leaders of Russia-based IT worker delegations and the intermediaries who converted their earnings, much of it in cryptocurrency, into revenue for the regime's weapons programs. The same day, a Virginia man who had worked at a U.S. defense contractor while holding an active Top Secret clearance pleaded guilty to lending his identity to the scheme. We come back to that case below.

The laptop farm is the critical piece of infrastructure. CrowdStrike, which helped the FBI disrupt the Texas operation, explains why: without a U.S. presence, the operatives couldn't receive and configure employer hardware, beat geolocation-based fraud controls, or get through the identity checks that guard legitimate employment. The farms keep adapting. CrowdStrike has seen operatives file fraudulent police reports claiming a laptop was stolen to trigger a replacement shipment to a new address, and route devices through IT staffing firms without the firm's knowledge. The FBI says the networks now run more laptop farms with fewer laptops at each, to avoid a single point of failure.

AI turned a scheme into an industry

Generative AI made the scheme much easier to scale. CrowdStrike reported a 220% year-over-year increase in activity from the group it tracks as Famous Chollima, which infiltrated more than 320 companies in a 12-month period. Operatives used generative AI to draft resumes, create false identities, mask their faces in video interviews, and answer questions or complete technical coding tests. On the job, they used it to handle daily work and juggle communications across as many as three or four jobs at once.

That trend continued into 2026. In CrowdStrike's 2026 technology threat landscape report, covering April 2025 through March 2026, Famous Chollima accounted for 47% of all state-sponsored hands-on-keyboard operations against the tech sector. In March, Microsoft reported that the North Korean group it tracks under the name Jasper Sleet uses generative AI to research job postings, tailor personas to in-demand skills, and swap operatives' faces into stolen identity documents. In September, a joint advisory from Japanese, U.S., Australian, and German agencies showed the same tools at work in WaterPlum, a related North Korean campaign in which operatives pose as recruiters: they ran online interviews with AI face-swapping software, then switched off video after a few minutes and blamed network issues.

The largest employers are seeing the pressure firsthand. Amazon has blocked more than 1,800 job applications from suspected North Korean operatives since April 2024, and it saw a 27% quarter-over-quarter rise in such applications in 2025. Applications for AI and machine learning roles drew particular interest, suggesting North Korea wants the roles closest to your most valuable intellectual property.

The network is getting broader and harder to see

The FBI's October announcement, titled "North Korean IT Workers Expanding Use of Facilitators, Broadening Recruitment," describes a scheme that spreads its risk across more people. Groups of North Korean IT worker teams are hiring Americans and others for a broader range of facilitation tasks. Some facilitators are paid only for the use of their identities. Others rent out their social media and professional networking accounts so operatives can apply for jobs behind an established profile. Operatives also subcontract the work itself to freelancers, usually in Pakistan, India, Nigeria, or neighboring countries.

Two findings should concern the defense community most. The FBI says these workers target positions with cleared defense contractors and other companies on U.S. government contracts, and that they approach active-duty U.S. military members to host laptops.

The Virginia case shows how close this has already come. According to the Justice Department, John Allen Kurrle Jr., 62, worked at a U.S. defense contractor from June to August 2025 while holding an active Top Secret clearance with SCI access. He let foreign individuals use his identity to apply for remote jobs, used a third person's identity without consent for others, and received the employers' laptops at his home so they could be operated from outside the country. The applications produced nine job offers from U.S. companies. He pleaded guilty to conspiracy to commit wire fraud on October 28 and faces up to 20 years in prison. A clearance vouches for the person who holds it, not for whoever is at the keyboard.

What they take

The paycheck funds the regime, but the access does more damage.

The FBI has observed North Korean IT workers using their access to company networks to steal proprietary and sensitive data, support cybercrime, and generate revenue for the regime. Workers have copied company code repositories, such as GitHub, to their own profiles and personal cloud accounts. The bureau also warns that they could collect company credentials and session cookies, use them to start work sessions from non-company devices, and open the door to further compromise.

For the defense industrial base, this has already happened. In June 2025, DOJ said North Korean workers had stolen employer data and source code, including International Traffic in Arms Regulations (ITAR) data, from a California-based defense contractor. In April 2026, two New Jersey men who ran laptop farms for that scheme were sentenced to 108 and 92 months in prison. Prosecutors said the operation used the identities of more than 80 Americans to land jobs at more than 100 U.S. companies and generated more than $5 million for North Korea. Along the way, it put export-controlled defense technical data in the hands of a sanctioned nuclear state.

For the contractor, it's also a compliance event. It increases ITAR exposure and triggers DFARS 252.204-7012 incident reporting. The CMMC timeline is in flux: in July 2026, the Pentagon suspended the Phase 2 requirement for third-party CMMC assessments, which had been set to begin November 10, pending a reform review. The underlying obligations haven't moved. DFARS 7012 and NIST SP 800-171 still apply, contractors still file self-assessments, and an incident like this still raises fair questions about whether the access controls were working in the first place. Google has seen the same pattern in Europe, where one operative ran at least 12 personas in late 2024 while applying for jobs at defense and government organizations.

Fire them, and they come back as extortionists

What happens when you catch one is just as important, because the threat doesn't end at termination.

The FBI reports that North Korean IT workers, once discovered, have held stolen data and code hostage until companies pay, and in some cases have publicly released proprietary code. Google's researchers assess that since late October 2024, these workers have increased the volume of extortion attempts and moved on to larger organizations, threatening to leak sensitive data, including source code for internal projects, or hand it to a competitor. September's joint advisory documented one such case, in which a North Korean IT worker extorted a company over payment and then published its proprietary source code online. As law enforcement pressure rises, the adversary is turning to tactics that look a lot like ransomware without the encryption.

It can get personal, too. The FBI has reported death threats against former managers and their families after workers were fired. Its October announcement adds that facilitators have made in-person threats against companies that removed operatives' accounts.

Security teams should treat offboarding a suspected operative as an incident. Revoke sessions and tokens, rotate secrets, audit repositories, and preserve evidence before anyone sends a termination email.

The code they write ships in your product

These operatives often do real work, and some do it well. Writing in Lawfare, Tom Uren notes that North Korean IT workers are sometimes exceptional performers and may be attempting to enable supply chain hacks.

A developer you can't trust may have committed code to your product, written your internal tooling, or configured your build pipeline. Because operatives now subcontract work to freelancers, the person who wrote a given commit may not be the person you hired, or anyone you've vetted. Each of those contributions is a potential foothold that outlives their employment. For a company that ships software into government or critical infrastructure environments, a fraudulent hire becomes a software supply chain problem for every customer downstream.

Insider IP theft isn't new

North Korea didn't invent the insider threat. American tech companies have dealt with trusted employees walking out with their most valuable IP for years. In 2022, former Twitter employee Ahmad Abouammo was convicted of passing personal data on Saudi government critics to Saudi officials.

Google's case is more recent and more instructive. Prosecutors showed that between May 2022 and April 2023, engineer Linwei Ding took more than 2,000 pages of Google's AI trade secrets and uploaded them to a personal cloud account. He told investors he could build an AI supercomputer by copying and modifying Google's technology.

How the case ended should give pause to anyone relying on prosecutors to protect their IP. A jury convicted Ding in January 2026 of economic espionage and trade secret theft. In August, the judge vacated the espionage counts, finding too little evidence that Ding meant to benefit the Chinese government. He was sentenced to just under a year in prison. At trial, his attorney argued that Google hadn't adequately protected the information and that it was widely accessible internally.

We think that defense argument is the most useful part of the story. Proving state intent in court is hard, and the company doesn't control that outcome. It does control who can reach its most sensitive IP.

As tech embraces defense, the target grows

Commercial tech is moving deeper into national security work, which raises the stakes.

In 2018 and 2019, the industry was pulling back. Thousands of Google employees protested the company's Project Maven drone-imagery contract, Google dropped out of bidding for the Pentagon's JEDI cloud contract, and Microsoft employees pushed back on a $479 million Army augmented-reality contract.

Since then, the industry has moved the other way. In February 2025, Google removed a pledge from its AI principles that had said it would not use the technology to develop weapons. In June 2025, the Army commissioned senior executives from Palantir, Meta, and OpenAI directly into the Army Reserve as lieutenant colonels through a new program called Detachment 201.

The same commercial developers, repositories, and cloud environments that once sat outside the defense perimeter feed directly into it. Every engineer's laptop at a company building AI or cloud capability for the military is part of the defense attack surface, and North Korea's interest in AI and machine learning roles suggests it understands that. That's why the hiring pipeline at a commercial tech company is now a national security control point.

Access matters more than citizenship

Some employers respond by screening out foreign names or treating every visa holder as a suspect. That approach is unfair, and it doesn't improve security. The FBI specifically cautions employers against relying only on ban lists or name-based screening, which can block qualified applicants and harm people whose identities were stolen. The North Korean scheme works because it borrows real American identities, and many of the facilitators in these cases were Americans. In the Ohio indictment, two of the people charged are U.S. nationals. Kurrle held a Top Secret clearance. Abouammo was a dual U.S.-Lebanese citizen. Ding had been a lawful permanent resident since 2021.

Nationality didn't predict any of these breaches, and neither did a clearance. In each case, the common factors were too much access and too little monitoring.

The good news is that you don't need to know who the adversary is. Control what any new hire can reach, confirm that the person you hired is the one doing the work, and notice quickly when either assumption breaks.

A defense playbook for the insider you didn't mean to hire

Look back at the cases above, and each one broke at a seam: an interview that trusted a video feed, a laptop that shipped without anyone confirming where it landed, a repository anyone on the team could clone, an offboarding that began with an email instead of an investigation. The FBI's October guidance is blunt about why. These workers "cannot be identified reliably with a single indicator." Its answer is what it calls "identity in depth," layered controls that run across the entire employment lifecycle.

The six steps below follow that lifecycle, from the first interview through the last day and out to the partners who supply your workforce. Each step needs a named owner. These schemes succeed in the gap between HR, which assumes security checked, and security, which assumes HR did.

1. Make identity verification a security control. Verify candidates live on camera against their documents, keep the camera on for the whole interview, and ask for unscripted actions that a face-swap overlay struggles to follow. Treat video that drops for "network issues" as a flag. North Korean recruiter scams already use that excuse to cut a face-swapped call short, and nothing stops a fake applicant from doing the same. Confirm that a new hire's equipment ships to the address they gave you, and look for reused contact details. According to the FBI, North Korean workers have used the same phone numbers and email addresses on resumes for what were presented as different applicants, and some now apply through rented professional networking accounts, so an established profile isn't proof of identity. The FBI also suggests adding verification friction, such as E-Verify, multiple forms of ID, or live verification. Amazon's approach is worth studying. It combines AI-powered screening with human verification and checks for ties to high-risk institutions, inconsistencies across applications, and location details that don't line up. Watch for requests to change the shipping address after the offer goes out, and for payroll sent to money transfer services or accounts in someone else's name. Owners: talent acquisition, payroll, and security.

2. Lock down the endpoint so a laptop farm can't use it. The FBI recommends least privilege: disable local administrator accounts and restrict who can install remote desktop applications. Hunt for remote-access tools nobody approved, such as AnyDesk, TeamViewer, RustDesk, Chrome Remote Desktop, and Splashtop, and for KVM-over-IP hardware such as PiKVM, TinyPilot, and GL.iNet devices. Flag software with no business purpose on the device, including OBS Studio with the VDO.Ninja plugin. Other warning signs include logins from residential networks that don't match where the employee says they live, a camera that stays off after onboarding, mouse jigglers, working hours that don't fit the stated time zone, and a laptop reported stolen with a request to ship the replacement somewhere new. Owners: IT and the SOC.

3. Protect the code and data themselves. The FBI recommends monitoring network logs and browser session activity for exfiltration using shared drives, cloud accounts, and private code repositories. Set alerts for repository clones to personal accounts. Segment access to source code, controlled unclassified information (CUI), and ITAR data so a new hire can't reach your most sensitive assets on day one. For contractors handling CUI, this falls under the access control and personnel security requirements in NIST SP 800-171, the same controls contractors attest to in their CMMC self-assessments. Owners: security and data owners.

4. Review code from new contributors before you trust it. Require code review, signed commits, and build pipeline integrity so a malicious change can't slip into production unnoticed. Watch for signs that more than one person is working a single account, such as abrupt shifts in coding style, commit times, or language, since the FBI says operatives subcontract their work to third-party freelancers. Owners: engineering and application security.

5. Plan the exit before you need it. Build a playbook for the day you discover an operative. It should cover session revocation, secret rotation, evidence preservation, reporting to your local FBI field office or ic3.gov, and the extortion attempt that may follow. Treasury has warned since 2022 that companies hiring or paying these workers face sanctions risk, and its October 2026 designations show that enforcement is still widening, so bring legal in early. Owners: security, legal, and HR.

6. Hold staffing partners and subcontractors to the same standard. In the defense industrial base, much of the workforce comes through third parties, and your controls are only as strong as theirs. CrowdStrike has seen operatives use IT staffing firms to redirect devices without the firm's knowledge. Put identity verification and device controls into staffing and subcontract agreements, require equipment to ship only to verified addresses, and ask for evidence that those controls work. Owners: procurement, supplier risk, and security.

How WWT can help

No single product stops a fraudulent hire. It takes HR, IT, security, legal, and procurement working from the same playbook, backed by technology proven to work together in your environment. WWT can help in three ways.

Find the seams before an adversary does. WWT's Executive Cyber Advisors work with CISOs and their peers in HR, legal, and procurement to map how hiring, onboarding, access, and offboarding controls hand off to one another, find the gaps an operative could slip through, and assign a clear owner to each step of the playbook, including the staffing partners and suppliers that feed a defense contractor's workforce.

Prove the controls before you deploy them. In WWT's Advanced Technology Center (ATC), a proving ground backed by more than 200 OEMs, your teams can evaluate identity verification, endpoint detection, data loss prevention, and insider risk tools side by side, and see how well they catch the behaviors described here, from unapproved remote-access tools and KVM-over-IP devices to repository exfiltration.

Rehearse the response. The WWT Cyber Range gives security teams a place to sharpen their detection and response skills, and our advisors can help you build and test the offboarding and extortion playbook before you need it.

The hiring pipeline is the front line

For decades, we built security around the perimeter with firewalls, gateways, and air gaps. North Korea has shown that the most reliable way in is through the front door, with an offer letter and a company laptop.

KnowBe4 caught its operative in minutes because someone was watching the endpoint. Organizations that build that kind of attention into how they hire, onboard, monitor, and offboard, with a named owner for each step, will protect their IP and their mission.