Cyber Threat Weekly – #117

The week of February 16^th through February 22^nd, about 322 cyber news articles were reviewed. A light-ish amount of cyber threat trends and adversarial behavior news to share.  Been thinkin about the effects of AI hype and the fear of missing out. 

It so happens several articles in this edition magnify the discussion around AI, just how the news cycle landed.  We see attackers embracing AI tech to scale and supercharge their own capabilities.  AI technology is enhanced every couple of months with new products launched consistently.  Organizations rush to roll out agentic AI, even with security lagging.  The rush to adopt AI without proper governance has a long-term cost when bad happens.

Let’s start with a low skilled attacker abuses AI to scale.  Claude Code Security launched by Anthropic.  2026 Agentic Coding Trends Report.  AI agents can and will go rogue.  BeyondTrust Remote Support exploitation surges.

Top-notch Starkiller phishing-as-a-service kit.  State of AI Security 2026.  2025 Cloud Threat Hunting and Defense Landscape.  Global Incident Response Report 2026.  Remote management and monitoring software abused extensively.

Broken Record Alert: Don’t get pwned by N-day vulnerabilities!!!

Known exploited software flaws are one of the top 4 initial access vectors and have increased sharply in recent months.  We continue to share n-day vulnerabilities being actively exploited.  Priority #1, start with the CISA / VulnCheck known exploited vulnerability (KEV) catalogs.  If it’s in the catalog, it should be patched.

A close #2 priority is flaws with weaponized proof of concept (PoC) code available.  Exploit chances are higher with weaponized PoC code available.  If you do nothing else with patching, have an emergency 24-to-48-hour patching process for actively exploited and weaponized PoC code available vulnerabilities.

VPN gateways from all vendors are under constant attack.

You should consider what is exposed to the Internet.  Architecture and zero trust network access (ZTNA) can go a long way to minimizing the number of devices and services exposed to the Internet.

CISA Known Exploited Vulnerabilities – February 16^th to February 22^nd:

CVE-2020-7796 – Synacor Zimbra Collaboration Suite (ZCS) Server-Side Request Forgery Vulnerability:
Contains a server-side request forgery vulnerability if WebEx zimlet installed and zimlet JSP is enabled.

CVE-2024-7694 – TeamT5 ThreatSonar Anti-Ransomware Unrestricted Upload of File with Dangerous Type Vulnerability:
ThreatSonar Anti-Ransomware does not properly validate the content of uploaded files. Remote attackers with administrator privileges on the product platform can upload malicious files, which can be used to execute arbitrary system commands on the server.

CVE-2008-0015 – Microsoft Windows Video ActiveX Control Remote Code Execution Vulnerability:
The bug could allow remote code execution. An attacker who successfully exploited this vulnerability could gain the same user rights as the logged-on user.

CVE-2026-2441 – Google Chromium CSS Use-After-Free Vulnerability:
Could allow a remote attacker to potentially exploit heap corruption via a crafted HTML page. This vulnerability could affect multiple web browsers that utilize Chromium, including, but not limited to, Google Chrome, Microsoft Edge, and Opera.

CVE-2021-22175 – GitLab Server-Side Request Forgery (SSRF) Vulnerability:
Contains a server-side request forgery (SSRF) vulnerability when requests to the internal network for webhooks are enabled.

CVE-2026-22769 – Dell RecoverPoint for Virtual Machines (RP4VMs) Use of Hard-coded Credentials Vulnerability:
Contains a use of hard-coded credentials vulnerability that could allow an unauthenticated remote attacker to gain unauthorized access to the underlying operating system and root-level persistence.

CVE-2025-49113 – RoundCube Webmail Deserialization of Untrusted Data Vulnerability:
Allows remote code execution by authenticated users because the _from parameter in a URL is not validated in program/actions/settings/upload.php.

CVE-2025-68461 – RoundCube Webmail Cross-site Scripting Vulnerability:
Contains a cross-site scripting vulnerability via the animate tag in an SVG document.

Low Skilled Attacker Breaches 600+ Fortinet Firewalls

To scale the campaign two large language models (LLMs) were used to assist the threat actor.  The LLMs generated attack methodologies, planned lateral movement strategies, and more.  Researchers discovered a server with tools and data piecing together the attack campaign.

https://www.bleepingcomputer.com/news/security/amazon-ai-assisted-hacker-breached-600-fortigate-firewalls-in-5-weeks/

https://aws.amazon.com/blogs/security/ai-augmented-threat-actor-accesses-fortigate-devices-at-scale/

https://cyberandramen.net/2026/02/21/llms-in-the-kill-chain-inside-a-custom-mcp-targeting-fortigate-devices-across-continents/

Claude Code Security – AI Vulnerability Scanning

Anthropic when they released Opus 4.6 shared, they had found 500 bugs in open-source code using the new model.  They have now released Claude Code Security, in a limited research preview.  It’s an AI bug scanner capable of discovering hard to find bugs and providing a proposed fix.  The double-edge sword is its dual use; attackers could benefit too.

https://thehackernews.com/2026/02/anthropic-launches-claude-code-security.html

https://www.anthropic.com/news/claude-code-security

2026 Agentic Coding Trends Report

This is a must-read report IMHO.  Anthropic is pattern matching and predicting trends for 2026.  Large language models are getting better with every release.  If you are paying attention to what’s happening with AI, these predictions make sense and are tied to outcomes their customers are getting currently.

https://claude.com/blog/eight-trends-defining-how-software-gets-built-in-2026

https://resources.anthropic.com/hubfs/2026%20Agentic%20Coding%20Trends%20Report.pdf?hsLang=en

Securing AI Agents is a Work in Progress

Understanding the nature of agentic AI is the key to securing AI agents.  Large language models are designed to reach their goals.  AI agents run on top of these models; they will discover and take advantage of any access they have to attain their goals.  Principle based security is more important than ever.

https://www.darkreading.com/application-security/ai-agents-ignore-security-policies

https://discuss.ai.google.dev/t/critical-bug-report-ai-studio-gemini-ignores-explicit-user-commands-overrides-safety-gates/109348

Critical Bug Exploitation in BeyondTrust Remote Support Ramping Up

This one is for tracking purposes and an update on activity.  Since the exploit code was released publicly, exploitation took off.  CISA even added exploitation by Ransomware to the known exploited vulnerability catalog.  Researchers share observations as they respond to victims. 

https://www.cybersecuritydive.com/news/beyondtrust-remote-support-exploitation-backdoors-remote-tools/812707/

https://unit42.paloaltonetworks.com/beyondtrust-cve-2026-1731/

Phishing-as-a-Service just Stepped Up – Starkiller

Solving many attacker problems, this kit is in its own category.  Starting with proxying the actual website through attacker-controlled infrastructure.  High-end reverse proxy tradecraft delivered in a simple workflow.  Once the victim enters their login credentials and even multi-factor code, they get into their account, but so does the attacker.

https://www.darkreading.com/threat-intelligence/starkiller-phishing-kit-mfa

https://abnormal.ai/blog/starkiller-phishing-kit

State of AI Security 2026

An interesting report detailing the AI supply chain as a huge attack surface and attack vector.  Rapid adoption of model context protocol and agent-to-agent protocols, current agentic AI standards, are just the start.  Lack of governance, also known as excessive agency, will be our achilles heal. 

https://www.cybersecuritydive.com/news/ai-agents-model-context-protocol-cisco-report/812580/

https://learn-cloudsecurity.cisco.com/2026-state-of-ai-security-report

2025 Cloud Threat Hunting and Defense Landscape

Researchers share trends of threat actors exploiting cloud infrastructure.  A few consistent patterns have been observed.  Initial access generally comes from Internet exposed services that are misconfigured or vulnerable.

https://www.recordedfuture.com/research/2025-cloud-threat-hunting-defense-landscape

Global Incident Response Report 2026

The attack on trust is observed and detailed in this report.  Identity played a compelling role in 90% of investigations.  AI is reducing the attack timeline.  Supply chain risk and trusted integrations were observed in 23% of investigations.  Exposure is beating sophistication; hence attacks continue to succeed.

https://www.csoonline.com/article/4133342/cyber-attacks-enabled-by-basic-failings-palo-alto-analysis-finds.html

https://www.paloaltonetworks.com/blog/2026/02/unit-42-global-ir-report/

https://www.paloaltonetworks.com/resources/research/unit-42-incident-response-report

Remote Monitoring and Management (RMM) Tool Abuse

A fresh report from Huntress, the 2026 Cyber Threat Report details the massive abuse of remote monitoring and management tools in cybercrime.  The abuse of legitimate tools and infrastructure is a continuing trend, and RMM tools are no exception.

https://www.darkreading.com/application-security/rmm-abuse-explodes-hackers-ditch-malware

https://www.huntress.com/press-release/huntress-cyber-threat-report-exposes-the-playbook-for-organized-cybercrime