{
 "$schema": "https://redlens.liminallayers.com/schemas/redlens.agents/3/edition.json",
 "schema_version": "redlens.agents/3",
 "date": "2026-10-21",
 "issue": 26,
 "status": "published",
 "incidents": [
  "RL-I-2026-0761",
  "RL-I-2026-0816",
  "RL-I-2026-0820",
  "RL-I-2026-0822"
 ],
 "removed": 0,
 "examined": 33,
 "window_until": "2026-10-21T16:33:45Z",
 "windowUntil": "2026-10-21T16:33:45Z",
 "summary": [
  "CrowdStrike details research into four threat actors who used AI agents to conduct agentic campaigns between May and August 2026 [1, 2]. The U.S. Department of Homeland Security has ordered federal civilian agencies to remove DeepSeek products from their systems and banned new purchases [2, 3]. Unit 42 reports that a threat actor used AI to plan and execute multiple types of cybercrime activities against targets in India [4]. Symantec researchers uncovered a malicious AI agent designed to hijack exposed large language model endpoints to run autonomous attacks [5]."
 ],
 "main": {
  "headline": "CrowdStrike Details Four Agentic AI Threat Actors",
  "paragraphs": [
   "CrowdStrike’s Counter Adversary Operations team published a report detailing research into four previously unreported threat actors who left their offensive agent frameworks exposed via open directories. The firm states that between May and August 2026, it documented incidents in which these actors used AI to conduct agentic campaigns in which AI agents make decisions and execute commands. CrowdStrike notes that these cases showcase threat actors delegating operational decision-making across the kill chain to agentic AI [1].",
   "The report describes a Spanish and English-speaking operator who compromised at least 15 Argentina-based healthcare entities using a dual-agent architecture derived from the open-source PentestGPT project. A Chinese-speaking actor targeted government entities in two countries and private sector firms in Brazil and the U.S., using Claude models via a China-based proxy service. This actor built two pentesting harnesses, one deterministic and one allowing Claude Code to operate as a single agent or delegate to multiple specialist agents in parallel [1].",
   "An Indonesian-speaking actor used 819 skills, primarily from public GitHub repositories, to instruct Claude Code to compromise targets. CrowdStrike notes this actor had to periodically intervene to overcome safety guardrails but still compromised several Indonesian government domains and private-sector entities. A Russian-speaking group exfiltrated data from a U.S.-based customer relationship management vendor and compromised entities in Russia, Kyrgyzstan, and the U.S. This group used a single slash command to task Claude Code with a seven-step workflow that granted near-total autonomy in selecting exploitation methods [1]."
  ],
  "paragraph_refs": [
   [
    1
   ],
   [
    1
   ],
   [
    1
   ]
  ]
 },
 "supplementals": [
  {
   "headline": "Unit 42 Details AI-Planned Cybercrime Campaign",
   "paragraphs": [
    "Unit 42 reports that a threat actor used AI assistants to plan and execute multiple types of cybercrime, including credential theft, VoIP fraud, and Wi-Fi password cracking. The actor, operating under the handle BUGSTER, configured the Hermes Agent framework with DeepSeek V4 as its reasoning engine and used Cursor as a secondary tool with safety controls disabled. Unit 42 states the actor used AI-generated scripts to conduct a brute-force campaign against an Indian university and exfiltrated a database of more than 5,200 records from a second university [4].",
    "The actor also developed VoIP fraud infrastructure using AI-generated advice on setting up fraudulent operations, including registering a U.S. limited liability company and bypassing caller ID authentication. Unit 42 notes that the actor’s conversations with jailbroken AI tools included step-by-step advice on these activities. The firm assesses that the actor lacked the skills these activities would have required before the advent of readily available AI tools [4]."
   ],
   "paragraph_refs": [
    [
     4
    ],
    [
     4
    ]
   ]
  },
  {
   "headline": "DHS Orders Removal of DeepSeek From Federal Systems",
   "paragraphs": [
    "The U.S. Department of Homeland Security has ordered federal civilian agencies to remove DeepSeek products from their information technology systems and banned new purchases. Homeland Security Secretary Markwayne Mullin issued the order under the Federal Acquisition Supply Chain Security Act of 2018, which was published in the Federal Register on October 20 [2]. The order names Hangzhou DeepSeek Artificial Intelligence Co., Ltd., and covers its affiliates, requiring agencies to purge systems within 60 days [2, 3, 6].",
    "The order cites concerns about data sovereignty, lax Chinese laws, and security vulnerabilities, stating that DeepSeek’s servers are located in the People's Republic of China. It also covers third-party products or services developed using DeepSeek implementations. The order does not apply to private consumers, the general public, state or local governments, or Department of Defense national security systems. Agencies can ask the official who issued an order for an exception, and under the council's regulations the request must be in writing and give a compelling justification [2, 3, 6]."
   ],
   "paragraph_refs": [
    [
     2,
     3,
     6
    ],
    [
     2,
     3,
     6
    ]
   ]
  },
  {
   "headline": "Symantec Finds AI Agent Hijacking Models for Attacks",
   "paragraphs": [
    "Symantec researchers report that they uncovered a malicious AI agent designed to run full cyber-attacks without human intervention. The agent was designed to use hijacked AI models to autonomously run up to 100,000 malicious actions, including scanning for targets, exploiting vulnerabilities, and deploying cryptocurrency miners. Symantec states that the researchers found the agent using an AI honeypot, which impersonates an unprotected, internet-facing AI service to attract and study attacks [5].",
    "The honeypot recorded over 127,000 external requests, and Symantec says the agent searches for exposed endpoints linked to large language models to use their computing power without paying for it. This technique, sometimes called LLMjacking, allows attackers to shift costs onto victims and gain anonymity. The captured instructions included about 74 tools for reconnaissance and exploitation, along with layered instructions designed to override safety controls. Symantec concludes that the activity marks a shift in how attackers are using AI, embedding it into offensive tools rather than trying to trick public chatbots [5]."
   ],
   "paragraph_refs": [
    [
     5
    ],
    [
     5
    ]
   ]
  }
 ],
 "references": [
  {
   "n": 1,
   "source": "crowdstrike",
   "title": "CrowdStrike Uncovers New Agentic AI Campaigns",
   "url": "https://www.crowdstrike.com/en-us/blog/crowdstrike-uncovers-new-agentic-ai-campaigns/",
   "published": "2026-10-21",
   "evidence_class": "threat_intel_report",
   "evidenceClass": "threat_intel_report",
   "incident_id": "RL-I-2026-0820",
   "incidentId": "RL-I-2026-0820",
   "archive_url": null,
   "archiveUrl": null,
   "primary": true,
   "standing_class": "threat_intel_report"
  },
  {
   "n": 2,
   "source": "thehackernews",
   "title": "U.S. Orders DeepSeek Removed From Federal Civilian Systems, Bans New Purchases",
   "url": "https://thehackernews.com/2026/10/us-orders-deepseek-removed-from-federal.html",
   "published": "2026-10-21",
   "evidence_class": "threat_intel_report",
   "evidenceClass": "threat_intel_report",
   "incident_id": "RL-I-2026-0761",
   "incidentId": "RL-I-2026-0761",
   "archive_url": "https://web.archive.org/web/20261021095048/https://thehackernews.com/2026/10/us-orders-deepseek-removed-from-federal.html",
   "archiveUrl": "https://web.archive.org/web/20261021095048/https://thehackernews.com/2026/10/us-orders-deepseek-removed-from-federal.html",
   "primary": false,
   "standing_class": "secondary_report"
  },
  {
   "n": 3,
   "source": "infoworld.com",
   "title": "US bans DeepSeek across federal agencies, putting AI supply chains under scrutiny | InfoWorld",
   "url": "https://infoworld.com/article/4238150/us-bans-deepseek-across-federal-agencies-putting-ai-supply-chains-under-scrutiny-3.html",
   "published": "2026-10-20",
   "evidence_class": "threat_intel_report",
   "evidenceClass": "threat_intel_report",
   "incident_id": "RL-I-2026-0761",
   "incidentId": "RL-I-2026-0761",
   "archive_url": "https://web.archive.org/web/20261021033033/https://infoworld.com/article/4238150/us-bans-deepseek-across-federal-agencies-putting-ai-supply-chains-under-scrutiny-3.html",
   "archiveUrl": "https://web.archive.org/web/20261021033033/https://infoworld.com/article/4238150/us-bans-deepseek-across-federal-agencies-putting-ai-supply-chains-under-scrutiny-3.html",
   "primary": false,
   "standing_class": "secondary_report"
  },
  {
   "n": 4,
   "source": "unit42",
   "title": "“Next Exploit Opportunities”: How a Threat Actor Used AI to Help Plan and Execute Multiple Types of Cybercrime Activities",
   "url": "https://unit42.paloaltonetworks.com/ai-assisted-cybercrime-analysis/",
   "published": "2026-10-21",
   "evidence_class": "threat_intel_report",
   "evidenceClass": "threat_intel_report",
   "incident_id": "RL-I-2026-0822",
   "incidentId": "RL-I-2026-0822",
   "archive_url": "https://web.archive.org/web/20261021143517/https://unit42.paloaltonetworks.com/ai-assisted-cybercrime-analysis/",
   "archiveUrl": "https://web.archive.org/web/20261021143517/https://unit42.paloaltonetworks.com/ai-assisted-cybercrime-analysis/",
   "primary": true,
   "standing_class": "threat_intel_report"
  },
  {
   "n": 5,
   "source": "infosecurity-magazine.com",
   "title": "Symantec Uncovers AI Agent Hijacking Models for Autonomous Attacks - Infosecurity Magazine",
   "url": "https://infosecurity-magazine.com/news/symantec-ai-agent-llmjacking",
   "published": "2026-10-21",
   "evidence_class": "threat_intel_report",
   "evidenceClass": "threat_intel_report",
   "incident_id": "RL-I-2026-0816",
   "incidentId": "RL-I-2026-0816",
   "archive_url": null,
   "archiveUrl": null,
   "primary": false,
   "standing_class": "secondary_report"
  },
  {
   "n": 6,
   "source": "cio.com",
   "title": "US bans DeepSeek across federal agencies, putting AI supply chains under scrutiny | CIO",
   "url": "https://cio.com/article/4238132/us-bans-deepseek-across-federal-agencies-putting-ai-supply-chains-under-scrutiny.html",
   "published": "2026-10-20",
   "evidence_class": "threat_intel_report",
   "evidenceClass": "threat_intel_report",
   "incident_id": "RL-I-2026-0761",
   "incidentId": "RL-I-2026-0761",
   "archive_url": "https://web.archive.org/web/20261021033307/https://cio.com/article/4238132/us-bans-deepseek-across-federal-agencies-putting-ai-supply-chains-under-scrutiny.html",
   "archiveUrl": "https://web.archive.org/web/20261021033307/https://cio.com/article/4238132/us-bans-deepseek-across-federal-agencies-putting-ai-supply-chains-under-scrutiny.html",
   "primary": false,
   "standing_class": "secondary_report"
  }
 ],
 "audio": null,
 "audio_bytes": null,
 "audio_seconds": null,
 "license": {
  "name": "CC BY 4.0",
  "url": "https://creativecommons.org/licenses/by/4.0/",
  "attribution": "The Red Lens, https://redlens.liminallayers.com/",
  "terms": "Use it for anything, commercial or not, with no permission needed; credit The Red Lens with a link. Covers what this site wrote (editions, summaries, labels, incident records, data files); linked articles belong to their publishers."
 }
}
