Microsoft: North Korean hackers run AI across hiring scams
Microsoft says North Korean hackers use AI for fake recruiters, AI-built resumes and trojanized private GitHub repos. How a small team vets hires and locks repos.
North Korean hackers now use AI at every step of the job scam, from both sides of the interview table. On October 29, Microsoft Threat Intelligence published a report on how DPRK groups put AI into their operations. Its call: adoption is high. Three groups use AI in all of their workflows. One group made more than 100 targeted PDF lures in two months, dozens of them minutes apart. For a small software shop that hires remote developers, or whose developers take recruiter calls, this is a direct risk. It is not a nation-state story you can ignore.
What actually happened
Per Microsoft's report, Microsoft puts each North Korean group on a four-stage scale of AI adoption:
- Operationalized: Jade Sleet, Coral Sleet and Jasper Sleet. AI is part of routine work across all of their operations.
- Accelerating: Diamond Sleet, Onyx Sleet, Pearl Sleet and Sapphire Sleet.
- Expanding: Emerald Sleet.
The specifics that matter for small businesses:
- Fake recruiters at speed. Between June 8 and June 12, Jade Sleet posed as recruiters for a fake "AI-powered blockchain intelligence company." It offered software engineering jobs and set up interviews. In the same window it ran similar campaigns through two more front companies.
- Fake candidates at scale. Jasper Sleet uses generative and agentic AI to build resumes and personas. It gets help on coding tests and keeps several fake workers' messages consistent. Microsoft also lists AI changes to profile photos, video and voice.
- Your private repos. Since April 2026, Coral Sleet has run a workflow that trojanizes private GitHub repositories. Microsoft says it was "highly likely" built with an AI development platform. Microsoft names Cursor as one tool these groups use for malware work.
- Agentic tools. Since late October, Onyx Sleet and Pearl Sleet have used ARTEX, an open-source agentic penetration testing framework.
Microsoft also says these operations are still human-directed, not autonomous. AI makes the same old attacks faster. It does not replace the operator.
Why it matters for your business
The interview is the attack surface. One lure tells your developer to clone a repo and run a "take-home test." A fake hire gets a laptop and repo access in week one. Both start in hiring, and AI has removed the typos and bad English that used to give them away.
Vet people on video, live. Do one unscripted video call with camera on, and ask questions that are not on the resume. Check that the candidate's profiles don't copy someone else's. Microsoft calls staffing agencies "the easiest avenue" for these workers. Ask your agency how it verifies identity.
Treat recruiter code as hostile. No developer runs a stranger's repo on a machine that holds your client credentials. Use a throwaway VM or a cloud sandbox.
Lock your private repos down. Require hardware keys or passkeys on GitHub. Remove contractor access the day the work ends. Turn on branch protection so no single account can push to main. A private repo is only as safe as the weakest account that can write to it.
Key takeaways
- Microsoft rates North Korean AI adoption as high; three groups use AI in all their workflows
- AI makes fake recruiters, fake candidates and malicious "coding tests" faster and more convincing
- Coral Sleet has trojanized private GitHub repos since April 2026
- Operations are still human-directed, so basic controls still work
- Do live video vetting, sandbox recruiter code, and enforce strong auth and branch protection
Who can push to your production repo right now? We audit repo access, set up branch protection and passkey-only sign-in, and build onboarding and offboarding checklists that take minutes, not a week. Book a repo access review.
Sources: Microsoft Threat Intelligence, Microsoft on Jasper Sleet IT workers.
- #security
- #north-korea
- #ai-threats
- #hiring
- #github
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
TestMu Kane CLI Exploration: AI writes tests with no spec
TestMu AI's Kane CLI Exploration agent maps a whole web app from one URL and turns it into tests, even with no spec. Here's how to use AI test generation safely.
Read itOpenAI Python SDK 3.42: agent compaction and scoped API keys
OpenAI Python SDK 3.42.0 adds agent context compaction, session review and per-request API key scoping. Upgrade multi-tenant code first, then tune compaction.
Read it