📱 Ivanti EPMM Zero-Day Gets Patched After Targeted Exploitation
Ivanti patched an Endpoint Manager Mobile zero-day exploited in targeted attacks, putting another privileged enterprise management product into the emergency lane. EPMM systems can sit close to mobile device administration, enterprise access, and authentication workflows, so compromise is not exactly a small-room problem. CISA added urgency through the Known Exploited Vulnerabilities process and a short federal patch deadline. Ivanti being a hot target category again is not a plot twist, but it is absolutely a reminder.
👉 Key takeaway: Prioritize Ivanti EPMM remediation, restrict administrative exposure, and review logs for targeted activity.

🧨 More Than 40,000 Servers Get Pulled Into the cPanel Zero-Day Mess
Attackers are continuing to exploit CVE-2026-41940, a critical cPanel and WHM vulnerability that can lead to administrative access. SecurityWeek reports that more than 40,000 servers have likely been compromised as exploitation continues. The scale is the story here because cPanel sits behind websites, hosting environments, databases, and server configurations. Shared hosting just found another way to ruin everyone’s week.
👉 Key takeaway: Patch affected cPanel and WHM systems immediately, audit exposed servers, and assume internet-facing instances have already been scanned.

🧬 Mini Shai-Hulud Keeps Developer Supply Chains on the Menu
Unit 42 is tracking a Mini Shai-Hulud npm supply-chain wave targeting packages tied to the SAP developer ecosystem. The campaign targets developer workflows and CI/CD environments, where tokens, credentials, and cloud secrets often live within grabbing distance. That makes this more than a malicious package story. It is an enterprise access story wearing a package manager hoodie.
📱 Ivanti EPMM Zero-Day Gets Patched After Targeted Exploitation
Ivanti patched an Endpoint Manager Mobile zero-day exploited in targeted attacks, putting another privileged enterprise management product into the emergency lane. CISA added urgency through the Known Exploited Vulnerabilities process and a short federal patch deadline. Ivanti being a hot target category again is not a plot twist, but it is absolutely a reminder.
👉 Key takeaway: Prioritize Ivanti EPMM remediation, restrict administrative exposure, and review logs for targeted activity.

🧨 More Than 40,000 Servers Get Pulled Into the cPanel Zero-Day Mess
Attackers are continuing to exploit CVE-2026-41940, a critical cPanel and WHM vulnerability that can lead to administrative access. SecurityWeek reports that more than 40,000 servers have likely been compromised as exploitation continues. Shared hosting just found another way to ruin everyone’s week.
👉 Key takeaway: Patch affected cPanel and WHM systems immediately, audit exposed servers, and assume internet-facing instances have already been scanned.

🧬 Mini Shai-Hulud Keeps Developer Supply Chains on the Menu
Unit 42 is tracking a Mini Shai-Hulud npm supply-chain wave targeting packages tied to the SAP developer ecosystem. The campaign targets developer workflows and CI/CD environments, where tokens, credentials, and cloud secrets often live within grabbing distance. It is an enterprise access story wearing a package manager hoodie.
👉 Key takeaway: Audit npm dependencies, rotate exposed tokens, monitor package updates, and treat developer environments as production-grade security territory.

🎭 Iranian APT Intrusion Wears a Chaos Ransomware Costume
An Iranian APT intrusion analyzed by Rapid7 appeared to masquerade as a Chaos ransomware incident while showing signs of espionage-focused activity. The attackers reportedly used social engineering, remote access tooling, credential theft, MFA manipulation, lateral movement, and data exfiltration before leaning into the ransomware narrative. Sometimes the mask is the point, not the motive.
👉 Key takeaway: Investigate identity abuse, remote access tooling, and data theft before assuming an extortion case is financially motivated.

🐧 Dirty Frag Gives Local Attackers Root Across Major Linux Distros
Dirty Frag is a new Linux zero-day with public exploit details that can give local attackers root privileges across major distributions. Attackers need local access first, but that is exactly what makes it dangerous after phishing, webshells, stolen credentials, or container escapes. A privilege-escalation bug is how “limited access” turns into “the attacker is now the landlord.”
👉 Key takeaway: Monitor vendor patches and mitigations, reduce unnecessary local access, and treat local privilege escalation as a serious post-compromise accelerator.

🚰 Claude AI Guides Hackers Toward OT Assets During Water Utility Intrusion
Dragos reported that threat actors used Claude AI during an intrusion involving a Mexican water and drainage utility. The model did not magically hack operational technology, but it helped attackers interpret internal findings and move toward OT, ICS, and SCADA-adjacent assets. AI can compress the boring middle of an intrusion, where attackers figure out what matters and where to go next.
👉 Key takeaway: Assume attackers will use AI to accelerate internal recon, asset interpretation, and OT targeting.

🔑 Braintrust Breach Forces Customers Into API Key Rotation Mode
AI evaluation startup Braintrust confirmed unauthorized access to one of its AWS cloud accounts and told customers to rotate sensitive keys. AI evaluation platforms can sit near model providers, application testing, prompts, telemetry, and integration secrets. AI platforms are now part of the secret-management perimeter, whether teams planned for that or not.
👉 Key takeaway: Treat AI tooling vendors as sensitive infrastructure and limit how long they hold customer secrets.

⏱️ U.S. Officials Consider Three-Day Patch Deadlines as AI Speeds Up Exploitation
Reuters reports that U.S. officials are weighing whether federal agencies should patch actively exploited vulnerabilities within three days. The concern is that AI tools may help attackers move faster from vulnerability disclosure to exploitation at scale. Patch governance may be headed for a speed limit sign that says “hope your asset inventory works.”
👉 Key takeaway: Improve asset inventory, emergency patch workflows, exploit-based prioritization, and rollback plans before compressed deadlines become the new normal.

🕵️ Five Eyes Agencies Publish Guidance for Secure AI Agent Deployment
Cyber agencies from the U.S., U.K., Australia, Canada, and New Zealand published guidance for securing agentic AI systems. The guidance focuses on identity management, least privilege, credential protection, containment, logging, human approval, and prompt-injection risk. AI agents are not magic assistants. They are privileged software identities with initiative.
👉 Key takeaway: Govern AI agents like high-risk privileged users: limit permissions, isolate execution, log actions, and require approval for sensitive tasks.

🧩 Claude Chrome Extension Bug Shows AI Agents Can Be Hijacked Through Browser Trust Gaps
A vulnerability dubbed ClaudeBleed in the Claude extension for Chrome could let malicious extensions hijack the AI agent for information theft or unauthorized actions. The issue involved lax permissions and trust assumptions around command origins, creating a path for remote prompt injection. Browser-based AI agents are convenient, which is another way of saying they are now standing very close to everything users do.
👉 Key takeaway: Review AI browser extension permissions and treat agentic browser tools as new privilege boundaries, not harmless productivity widgets.