---
title: Darktrace turns Claude Code into a live C2 implant using stolen Claude Console API keys and the platform's memory stores, with all traffic going to api.anthropic.com | ctxsec
url: https://ctxsec.com/items/darktrace-claude-memory-stores-covert-c2-stolen-console-keys-claude-code-implant-lab
description: "In a lab, Darktrace Signal Labs showed that an attacker who already controls a device and has the user's Claude Console API credentials can create a memory store in the user's workspace and use three memory files (heartbeat, input, output) as a command channel. A systemd-run service repeatedly launched non-interactive Claude Code processes that checked in, picked up natural-language jobs and posted results. Across nine jobs the agent scanned the network, pulled names and a password from an anonymous FTP share, generated usernames, sprayed the password at the domain controller, listed and downloaded a 1 GB SMB file, and exfiltrated it in three parts via the Claude Files API. Darktrace says no Claude vulnerability is involved and it hasn't seen this in the wild."
status: "vendor research (Darktrace Signal Labs, Nov 12, 2026); lab demonstration, not observed in the wild, not a Claude vulnerability"
published: 2026-11-14
---

> In a lab, Darktrace Signal Labs showed that an attacker who already controls a device and has the user's Claude Console API credentials can create a memory store in the user's workspace and use three memory files (heartbeat, input, output) as a command channel. A systemd-run service repeatedly launched non-interactive Claude Code processes that checked in, picked up natural-language jobs and posted results. Across nine jobs the agent scanned the network, pulled names and a password from an anonymous FTP share, generated usernames, sprayed the password at the domain controller, listed and downloaded a 1 GB SMB file, and exfiltrated it in three parts via the Claude Files API. Darktrace says no Claude vulnerability is involved and it hasn't seen this in the wild.

[← Context & Harness Attacks](/context) · [markdown](/items/darktrace-claude-memory-stores-covert-c2-stolen-console-keys-claude-code-implant-lab.md)

vendor research (Darktrace Signal Labs, Nov 12, 2026); lab demonstration, not observed in the wild, not a Claude vulnerability

2026-11-14 7:35 PM ET

### Darktrace turns Claude Code into a live C2 implant using stolen Claude Console API keys and the platform's memory stores, with all traffic going to api.anthropic.com

- **Why it matters:** the C2 and exfiltration channel is the same endpoint the agent uses for inference, so network allowlists for AI tooling don't separate it.
- **Credential scope:** per Darktrace (to Help Net Security), standard API keys reach memory stores while OAuth credentials don't, so prefer OAuth sign-in for developer machines and keep Console API keys off endpoints.
- **Detection:** correlate agent harness processes with the processes they spawn and their network activity, and baseline what agents normally do (see the hunt).
- **Disclosure:** Darktrace shared the findings with Anthropic before publishing; there was nothing to patch.

Sources: [Darktrace: Living off the AI: How Claude Memory Stores Enable Covert C2 Channels (Nov 12, 2026)](https://www.darktrace.com/blog/living-off-the-ai-shared-memories-as-a-covert-c2-channel) · [Help Net Security: Stolen Claude API keys can turn Claude Code into a remote-controlled implant (Nov 12, 2026)](https://www.helpnetsecurity.com/2026/11/12/stolen-claude-api-keys-remote-controlled-implant/)
