---
id: "CMP-202611-cylyny"
page: "https://infrastack.ai/live/CMP-202611-cylyny"
kind: "campaign"
aliases: []
title: "North Korea-linked UNC6836 targets crypto firms with fake recruiters, then changes Claude Code settings so every launch downloads the SOMBERMEME backdoor (Google GTIG)"
severity: "high"
cvss: null
epss: null
exploited: true
status: "open"
category: "ai-tooling"
products: [{"name": "Claude Code", "fixed": [], "affected": "settings tampered after compromise (not a Claude Code bug)", "ecosystem": "npm"}]
ecosystems: []
tags: ["campaign", "early-warning"]
not_affected_if: []
published_at: null
first_seen_at: "2026-11-16T14:37:01+00:00"
updated_at: "2026-11-16T14:38:03+00:00"
sources: [{"source": "analyst", "url": "https://www.theregister.com/research/2026/11/16/google-ai-agent-turned-a-security-patch-into-a-working-exploit-with-no-human-help/5309926", "first_seen_at": "2026-11-16T14:37:01+00:00", "cites": ["https://www.theregister.com/research/2026/11/16/google-ai-agent-turned-a-security-patch-into-a-working-exploit-with-no-human-help/5309926"]}]
exploit_sources: []
---

# North Korea-linked UNC6836 targets crypto firms with fake recruiters, then changes Claude Code settings so every launch downloads the SOMBERMEME backdoor (Google GTIG)

Link for people: https://infrastack.ai/live/CMP-202611-cylyny (this .md is for agents)

Google Threat Intelligence Group's Q4 2026 AI Threat Tracker (16 Nov 2026) reports several attacks by UNC6836 on cryptocurrency firms in North America and Europe. GTIG suspects UNC6836 is linked to Midnight Neptune (formerly UNC1069, North Korea). It approaches targets with AI-generated fake recruiter profiles (ITmedia). After getting in, it changes the victim's Claude Code settings so that Claude Code downloads the SOMBERMEME backdoor every time it starts, which keeps the attacker in. GTIG's September 2026 report already described Midnight Neptune poisoning Claude CLI hooks with SOMBERMEME. This is new activity under a new cluster. Victims, dates and indicators have not been published in the coverage so far.

_Summary, check and fix written by a model from the sources; labelled as such._

## Affected

- npm:Claude Code affected settings tampered after compromise (not a Claude Code bug) fixed no fixed version listed

## Cause

How the attacker got in: a person was tricked (social-engineering). UNC6836 approaches targets with AI-generated fake recruiter profiles and the lure leads staff to run code, so the entry is a person being tricked, while the backdoor stays on the victims' own developer machines.

How it reached others: it stayed inside (none).

So it does not happen to you: Use phishing-resistant second factors on the accounts that matter, and never reset a factor on a phone call alone.

Last month's incidents with the same cause, and the full checklist: https://infrastack.ai/live/breaches.md#social-engineering

## How to check

- On developer machines (especially at crypto firms), check Claude Code settings files (~/.claude/settings.json, project .claude/settings.json and .claude/settings.local.json) for hooks or startup commands you did not add, especially any that run curl, wget, PowerShell or a script that downloads something.\n- Look for unexpected network connections or new processes started by the Claude Code process at launch.\n- Be suspicious of recruiter approaches that lead to running code or tools from the 'recruiter'.\n- SOMBERMEME hashes and C2: not published in the coverage; unknown.

## Fix

- Remove any Claude Code hooks or settings you cannot explain, and manage Claude Code settings centrally (managed settings) where you can.\n- If you find a malicious hook, treat the machine as compromised: isolate it, reimage it, and rotate the credentials, wallet keys and tokens that were on it.\n- Train staff at crypto firms on fake recruiter lures and never run code from a job approach on a work machine.

## Timeline

- 2026-11-16T14:37:01+00:00 analyst https://www.theregister.com/research/2026/11/16/google-ai-agent-turned-a-security-patch-into-a-working-exploit-with-no-human-help/5309926

## Documents

- [describes] Google warns that fully-automated AI attacks are coming soon https://www.bleepingcomputer.com/news/security/google-warns-that-fully-automated-ai-attacks-are-coming-soon/
- [describes] Google: AI agent turned a security patch into a working exploit with no human help https://www.theregister.com/research/2026/11/16/google-ai-agent-turned-a-security-patch-into-a-working-exploit-with-no-human-help/5309926
- [discussion] @GoogleCloudSec: New GTIG AI Threat Tracker:

As 2027 nears, human latency is leaving the attack  https://x.com/GoogleCloudSec/status/2122369113594966244

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_
