Opening

The FBI issued an advisory today (November 13, 2026) warning that the open-source penetration testing framework ARTEX has been weaponized by multiple threat actors, with attackers able to gain high-level privileges in as little as 3 hours. The impact spans finance, government, and defense sectors. Today we'll also cover an already-exploited management vulnerability in GPUStack that CISA is requiring patched, the investigation findings from the Cursor supply chain incident, and details on a North Korean deepfake scam seizure operation. I'm Muyan.

Today's AI Security Highlights

1. GPUStack unauthenticated management vulnerability actively exploited, patch deadline has passed

  • What happened: The AI inference platform GPUStack's CVE-2026-102895 (CVE is the industry's standardized vulnerability tracking numbering system) scored a 9.8 on CVSS (CVSS is a vulnerability severity scoring system, where higher scores indicate greater severity). Security research firm Ostorlab disclosed the flaw on November 12, 2026, noting that attackers can bypass authentication to gain unauthorized admin access and execute arbitrary code. CISA has added it to the Known Exploited Vulnerabilities (KEV) catalog and required remediation by November 12, 2026, a deadline that has already passed.
  • Who's affected: Any team managing AI inference services with GPUStack should take note; attackers have already been observed exploiting this vulnerability to gain host control in the wild.
  • Recommended action: Check your current version immediately and upgrade to the officially recommended 2.2.0 or later. If you haven't patched yet, first verify whether your admin interface is directly exposed to the public internet.
  • Source: Ostorlab (https://blog.ostorlab.co/gpustack-auth-bypass-rce-cve-2026-102895.html)

2. FBI warns: open-source AI pentest framework ARTEX weaponized, intrusions achieve full compromise in as little as 3 hours

  • What happened: On November 13, 2026, the FBI issued a FLASH advisory (FLASH-20261113-001) stating that the open-source multi-agent (multiple AI agents working together to complete tasks) penetration testing framework ARTEX has been abused by multiple threat actors since August 2026 to automate entire attack chains, from reconnaissance to data exfiltration. According to reports, some intrusions went from initial access to high-privilege compromise in under 3 hours.
  • Who's affected: The advisory notes impacts across finance, government, technology, defense, healthcare, and other sectors, showing that this kind of automated attack tool no longer discriminates by industry.
  • Recommended action: Focus detection efforts on "rapid, multi-step automated behavior." Strengthen EDR (Endpoint Detection and Response) sensitivity for alerts on abnormal privilege escalation and lateral movement, and reassess whether your incident response process can actually be activated within a 3-hour window.
  • Source: BleepingComputer (https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/)

3. North Korea's FaceStealer deepfake video scam nets over $400 million, US DOJ seizes infrastructure

  • What happened: On November 12, 2026, the US Department of Justice announced the seizure of cloud infrastructure used by North Korea's Sapphire Sleet group in its FaceStealer social engineering operation. Investigators found that over 18 months, the operation used deepfakes (AI-synthesized fake faces and voices) in video calls to impersonate well-known figures, netting over 400millioninfraud.Underafederalsearchwarrant,theFBIseizedmorethan900videosofvictims,andalsoobtainedauthorizationtoseize400 million in fraud. Under a federal search warrant, the FBI seized more than 900 videos of victims, and also obtained authorization to seize 1.2 billion in stolen funds linked to the Binance heist.
  • Who's affected: These attacks primarily target cryptocurrency industry professionals and high-net-worth individuals, building trust through video calls before inducing transfers or extracting credentials. Corporate finance teams and executives should stay vigilant.
  • Recommended action: For any video call requesting urgent transfers or credential changes, always verify identity through a second channel (phone call, known contact method); don't rely solely on the video call itself to judge authenticity.
  • Source: US Department of Justice (https://www.justice.gov/opa/pr/justice-department-announces-actions-disrupt-north-koreas-massive-theft-cryptocurrency)

4. Cursor confirms Fireworks AI key leak, malicious fake Claude Code extension sat on marketplace for over 11 hours

  • What happened: On November 13, 2026, Cursor published its root cause investigation into the October 29, 2026 incident, confirming that a leaked Fireworks AI access key was used to steal code agent conversation logs containing user credentials. Attackers then used these stolen credentials to inject a counterfeit Claude Code extension into the marketplace cache, which remained exposed for roughly 11 hours.
  • Who's affected: Cursor says up to 5,000 users were affected; the company has rotated the relevant credentials and rebuilt its marketplace security mechanisms.
  • Recommended action: If you installed a Claude Code extension around October 29, 2026, verify its source and rotate related API keys. More broadly, when using third-party AI service keys in daily work, regularly rotate them and limit their access scope, which reduces the damage from this type of supply chain incident (indirect compromise through upstream services or packages).
  • Source: The Hacker News (https://thehackernews.com/2026/11/stolen-fireworks-ai-key-led-to-cursor.html)

5. Unit 42 exposes CAIRN framework: autonomous AI agents targeting semiconductor, defense, and civil society organizations

  • What happened: On November 12, 2026, Palo Alto Networks' Unit 42 research team published a report revealing that the open-source autonomous AI agent (AI systems capable of independently planning and executing multi-step tasks) framework CAIRN has been used by attackers to carry out multi-stage intrusions against organizations in semiconductors, defense, and civil society.
  • Who's affected: Researchers believe this shows agentic attack tools continuing to spread into more industries, with civil society organizations, think tanks, and human rights groups, groups previously rarely considered high-value targets, now appearing on attacker target lists.
  • Recommended action: These organizations should prioritize auditing least-privilege settings on externally facing services and watch for unusual but seemingly "logical" sequences of multi-step operations, which are often a signature of automated agent-driven attacks.
  • Source: Palo Alto Networks Unit 42 (https://unit42.paloaltonetworks.com/cairn-framework-ai-autonomous-attacks/)

6. Taiwan's Administration for Cyber Security: AI-driven cyberattacks a growing trend, roughly 8 million personal records affected

  • What happened: On November 12, 2026, Taiwan's Ministry of Digital Affairs' Administration for Cyber Security reported that more than a dozen domestic companies and online platforms have reported AI-related cyberattacks, affecting an estimated 8 million personal records in total. Preliminary analysis suggests nearly half of recent security incidents involve AI-based attack techniques.
  • Who's affected: Companies of all sizes in Taiwan should recognize that AI-driven attacks are no longer just overseas case studies, they're now showing up directly in domestic incident reports.
  • Recommended action: Companies should reassess whether their personal data inventory and breach notification processes can keep pace with the speed of AI-driven attacks, and evaluate whether their current monitoring systems can distinguish AI-generated anomalous behavior patterns.
  • Source: Central News Agency (https://www.cna.com.tw/news/aipl/202611120178.aspx)

Closing

Of today's six stories, the GPUStack patch deadline has already passed, and open-source AI attack frameworks like ARTEX and CAIRN are rapidly spreading across industries. The key takeaway: the speed of automated attacks has compressed down to a matter of hours. When you get back today, check one thing first: is your team's key rotation schedule for AI-related tools and extensions still stuck in manual, "whenever we remember" mode? I'm Muyan, see you tomorrow.

Author

Mark Ku

10 年以上的軟體工程師,做過北美電商與 AI SaaS 訂閱收費系統。Read More

Found this useful?

The author's free tools, daily podcasts and newsletter are all here.

Mark Ku · This article is licensed under CC BY 4.0. Credit the author and link back to the original when reusing it.

Comments

Subscribe to Newsletter

Subscribe to get new posts delivered instantly — never miss a tech share.

By submitting, you agree to receive emails. You can anytime.

Popular Posts

View all
Mark Ku
··633

Oracle Cloud Always Free Tier: Linux Host and Static IP for a $0 Cloud Solution

Oracle Cloud Always Free Tier: Linux Host and Static IP for a $0 Cloud Solution
Mark Ku
··464

Say Goodbye to Postman's Fee Trap! A Hands-on Guide to Bruno, the Open-Source Git-Native API Testing Powerhouse.

Say Goodbye to Postman's Fee Trap! A Hands-on Guide to Bruno, the Open-Source Git-Native API Testing Powerhouse.
Mark Ku
··249

A Free, Open-Source, Notion-like Knowledge Base — A Complete Guide to Deploying and Backing Up Outline Wiki

A Free, Open-Source, Notion-like Knowledge Base — A Complete Guide to Deploying and Backing Up Outline Wiki
Mark Ku
··205

Building an Efficient API Management Platform: Deploying Kong Gateway from Scratch - Part 1

Building an Efficient API Management Platform: Deploying Kong Gateway from Scratch - Part 1
Mark Ku
··202

Setting Up Samba on Ubuntu to Share Folders with Windows 11

Setting Up Samba on Ubuntu to Share Folders with Windows 11
Mark Ku
··198

Training Your Own AI Voice: Hardware Requirements, Open-Source Model Comparison, and LoRA Fine-Tuning

Training Your Own AI Voice: Hardware Requirements, Open-Source Model Comparison, and LoRA Fine-Tuning