AI Job Scams in 2026: The Take-Home Assignment That Steals Your Credentials

If you're an AI or ML engineer, there is a scam running right now that is built precisely for you, and it works because it asks you to do something completely normal.

A recruiter emails about a role. The company is real — you can look it up. The recruiter's LinkedIn profile looks fine. The message is well written and specific to your background. They send a take-home coding assignment: a GitHub or GitLab repository. Clone it, open it in VS Code or Cursor, have a look at the task.

You do it, because that is what a take-home assignment is.

Opening that folder executes code. In one campaign documented in April–May 2026, the payload included three custom modules: one stealing saved credentials from Chrome and Firefox, one hunting for and exfiltrating crypto wallets, and one cleaning up forensic traces afterward.

Over six weeks, that single campaign sent more than 250 emails to nearly 100 companies. The recruiters impersonated real firms — a DeFi company, a telehealth startup, a Web3 and AI talent agency. Security researchers tracking it link it to a longer-running operation the industry calls Contagious Interview, attributed to North Korean state actors.

Nothing in ordinary job-search advice prepares you for that. So let's fix it.

---

The scale, in government numbers

Vendor statistics dominate most hiring articles. These are federal, which makes them worth more.

The FTC reports job scam losses reached $501 million in 2025 — up 457% since 2020. Separately, scam losses originating on social media hit $2.1 billion in 2025, eight times the 2020 figure.

The FBI's IC3 logged 20,044 employment fraud complaints in a single year, a 276% jump from 2023. On AI-enabled scams generally, the Bureau received nearly 22,000 complaints with $893 million in losses.

And the number that should reframe all of the above: the FTC estimates fewer than 10% of fraud victims report to any federal agency. The real totals are likely several times these figures.

Industry tracking puts the rise in deepfake hiring fraud attempts at roughly 1,300% between 2023 and 2024. Task scams — where you're paid small amounts to complete repetitive online tasks, then asked to deposit money to unlock higher-paying work — accounted for nearly 40% of job scam reports in the first half of 2024.

This is not a fringe risk anymore. It's a routine hazard of looking for work.

The five scams aimed at technical candidates

1. The malicious take-home

How it works: a plausible recruiter, a real company name, a repository you're asked to clone and open locally. Execution happens on open — via project configuration files, dependency install scripts, or editor tasks — before you've read a line of the actual assignment.

Why it works on us specifically: cloning a repo and opening it in an editor is muscle memory. It's the least suspicious action in a developer's day. The scam doesn't ask you to do anything unusual; it just makes the normal thing dangerous.

How to defend:

• Never open an unvetted repository directly on your main machine. Read the code on the web interface first, in the browser.
• If you need to run it, use a container or a throwaway VM with no credentials, no wallet, no SSH keys, no cloud config mounted.
• Disable automatic task execution in your editor. Both VS Code and Cursor can be configured not to auto-run workspace tasks and to prompt before trusting a folder. Do this now, not after.
• Be sharply more suspicious if the role touches crypto or fintech — that's where these campaigns concentrate.
• A legitimate company is completely fine with you saying: "I'll review this in a sandbox — can you also share the assignment as a document?" A refusal is your answer.

2. The deepfake interview

Real-time face-swap filters let a scammer impersonate a named executive on video. It cuts both ways: candidates get interviewed by people who don't exist, and companies interview candidates who aren't who they appear to be.

Signals: poor lip sync under fast speech, edge artifacts around hair and glasses, unwillingness to turn their head in profile, audio that doesn't match mouth movement, refusal to switch to a different platform mid-call.

Defense: insist on a call scheduled through the company's own domain and calendar system. Ask a question that requires specific, verifiable knowledge of the company's product — not something a script covers. If anything feels wrong, end the call and email the company through its published address.

3. The equipment and onboarding fee

You're hired. Now you need to buy a laptop, software licenses, or a "home office setup" — reimbursed on your first paycheck. Or they send a check to buy equipment, it clears provisionally, you spend it, and it bounces two weeks later leaving you liable.

The absolute rule: a legitimate employer never asks you to pay for anything to start work, and never sends you money to spend on their behalf before you're on payroll. There is no exception to this. If you find yourself constructing one, that's the scam working.

4. The credential and identity harvest

An "onboarding portal" collects your government ID, bank details, and tax forms before any employment relationship exists. Or a "background check" that asks for your Social Security number, date of birth, and mother's maiden name over email.

Defense: real employers collect this after a signed written offer, through a named HR platform, on a domain you can verify. Before that, they need your name, email, and résumé — nothing else. Not your ID, not your bank details, not your SSN.

5. The unpaid "trial project"

Not always criminal, but the same extraction: a take-home that is transparently real production work — build this specific feature we need, on our specific stack, with our specific data.

The test: a legitimate assessment is scoped to a few hours, is deliberately artificial, and has no direct commercial value. If you could deploy your submission and it would work in their product, you're being asked to work for free.

The general red flags

Fast checklist, in rough order of reliability:

• Contact from a free email domain — @gmail.com, or a lookalike like @anthropic-careers.com
• Immediate move to WhatsApp, Telegram, or Signal. Legitimate recruiting happens on email and named platforms. This is one of the strongest single signals.
• Hired with no real interview. An offer after one chat message is not a compliment.
• Any request for money, in any direction
• Compensation well above market for the seniority described
• Pressure and artificial urgency — "we need your details today to secure the role"
• A job description with details that don't match the company's actual product
• No verifiable web presence for the recruiter beyond a profile created recently
• Crypto payment, or payment in anything other than normal payroll
• Grammar and formatting that's too polished but subtly off-topic — AI-generated outreach at scale

How to verify a recruiter in five minutes

Do this before you reply to anything, not before you accept.

1. Find the job on the company's own careers page. Not a job board, not the link they sent — the company's own domain. If it isn't there, ask why.
2. Check the email domain character by character. nvidia.com versus nvidla.com. Homograph and lookalike domains are the whole trick.
3. Look up the recruiter independently. Search their name plus the company. Check profile creation date and connection count — a recruiter at a real company with 40 connections and a three-week-old profile is not a recruiter.
4. Contact the company through a channel they control. Their published careers email, or a current employee. "Is [name] a recruiter with you, and is this req real?" takes one message.
5. Search the exact text of the message. Scam templates get reused. Someone has usually posted it.

That's five minutes, and it defeats nearly every version of this.

If you've already been hit

If you cloned and opened a suspicious repository:
Assume compromise; act accordingly. Disconnect the machine from the network. From a different, clean device, rotate every credential stored in your browser — starting with email, then cloud providers, then source control. Revoke active sessions and API tokens. Enable or re-enroll hardware MFA. Move any crypto to a new wallet with new keys generated on a clean machine. Then get the affected machine examined properly rather than assuming a reinstall settled it — and if it's a work machine, tell your employer's security team immediately, today, even though it's uncomfortable.

If you sent money: contact your bank or card issuer immediately and use the words "fraud" and "chargeback." Speed matters more than anything else.

If you sent identity documents: freeze your credit with all three bureaus, and file an identity-theft report at IdentityTheft.gov.

Report it either way. ReportFraud.ftc.gov and the FBI's IC3.gov. Given that under 10% of victims report, your report is doing more work than you think — enforcement priorities follow complaint volume.

And don't absorb the shame. These operations are run by well-resourced, state-backed teams whose entire job is this. Falling for a well-built attack is not a character flaw, and the people who don't report are the reason it keeps scaling.

For hiring teams: the mirror image

There's a second version of this problem pointed at employers.

Federal investigations and reporting through 2026 have documented North Korean IT workers obtaining remote positions at US companies at scale, generating revenue that flows back to a sanctioned state — often with US-based facilitators, sometimes unwitting, providing addresses, identities, and laptop farms. Related campaigns recruited freelancers in the Philippines, Nigeria, Colombia, and Bangladesh to supply a convincing human face for interviews.

For a company, this is not merely a bad hire. It's a potential sanctions violation and a live insider-access problem.

Practical mitigations that don't degrade your candidate experience:

• Verify identity documents against a live video presence, at least once, with the camera on and the person willing to turn their head
• Confirm the work location matches the tax and payroll paperwork
• Watch for the known operational tells: repeated equipment-shipping address changes, a persistent refusal to appear on camera, timezone behavior inconsistent with the stated location, and payment-routing requests to third parties
• Be deliberate about when a new hire receives production credentials

And do not solve this with AI detection on résumés. It doesn't work, it misfires badly on non-native English speakers, and it addresses a different problem entirely. Identity verification is the control here; stylometry is not.

FAQ

How common are job scams in 2026?
The FTC recorded $501 million in job scam losses in 2025, up 457% since 2020. The FBI's IC3 logged 20,044 employment fraud complaints in a year, up 276% from 2023 — and the FTC estimates fewer than 10% of victims report at all.

Can a take-home coding assignment give me malware?
Yes, and it's an active campaign specifically targeting developers. Cloning a repository and opening it in VS Code or Cursor can execute code before you read the assignment. Review unknown repos in the browser first, and run them only in a container or throwaway VM with no credentials present.

What's the single biggest red flag?
A push to move the conversation to WhatsApp, Telegram or Signal early. Legitimate recruiting stays on email and named platforms.

Will a real employer ever ask me to pay for equipment?
No. Never, in any form, including "we'll reimburse you" and "here's a check to buy it." This rule has no exceptions.

How do I verify a recruiter is real?
Find the role on the company's own careers page, check the email domain character by character, look up the recruiter independently including their profile's age, and contact the company through a channel it controls.

I already opened a suspicious repo. What now?
Treat it as a compromise. Disconnect that machine, rotate all browser-stored credentials from a clean device, revoke sessions and tokens, re-enroll MFA, move any crypto to newly generated keys, and get the machine properly examined. Tell your employer's security team if it's a work device.

Are AI-generated job postings a real problem?
Yes — synthetic company profiles, AI-written postings, and AI-generated recruiter personas are all now cheap and common. Which is exactly why verification has to run against the company's own domain rather than against how professional the message looks.

---

The short version

The scam that will get you is not a badly spelled email promising $300,000 to work from home. It's a well-written message from a real company's name, about a role that fits your background, asking you to do the most ordinary thing a developer does.

So build one habit: verify through the company's own domain before you engage, and never execute an unknown repository on a machine that holds anything you care about. Those two rules stop nearly all of it.

And if you're already caught — rotate, report, and skip the shame. The operations doing this are professional. Being targeted says nothing about you except that you're employable.

---

DevFound is an AI-first job platform for AI and ML talent, with roles from verified AI-native companies. Browse open AI roles.

---

Sources

• Federal Trade Commission — job scam losses of $501M in 2025, up 457% since 2020; social-media-originated scam losses of $2.1B in 2025 (reported April 2026); estimate that fewer than 10% of fraud victims report to a federal agency
• FBI Internet Crime Complaint Center (IC3) — 20,044 employment fraud complaints in a single year, +276% versus 2023; ~22,000 complaints and $893M in losses relating to AI-enabled scams
• Proofpoint / security industry reporting on UNK_DeadDrop and Contagious Interview, April–May 2026 — 250+ emails to ~100 companies over six weeks; recruiter impersonation of real firms; malicious repositories executing on open in VS Code and Cursor; browserlogin, companywallet and cleanup modules; concentration on cryptocurrency-adjacent targets
• Reporting on North Korean remote IT worker schemes (NBC News investigation with Nisos and the FBI; Fortune; Wilson Sonsini and Crowell & Moring client alerts) — placement of workers in US companies at scale, US-based facilitators, and recruitment of freelancers in the Philippines, Nigeria, Colombia and Bangladesh
• Industry estimates of a ~1,300% rise in deepfake hiring fraud attempts, 2023–2024, and task scams comprising ~40% of job scam reports in H1 2024 — vendor-sourced, directional

The FTC and FBI figures are the load-bearing evidence here and are non-vendor. Attack-campaign details come from published security research; attribution to state actors is as reported by those researchers, not independently verified here.