IT Security News: Overnight roundup, 2026-11-18
Summary
Security news from the 00:00–06:00 CET window was dominated by vulnerabilities that can be exploited with little or no human involvement, with AI playing a role on both the attacking and defending sides. The most serious thread is a critical unauthenticated flaw in OpenJDK's Base64 decoder, while a supply-chain-style compromise of Optimizely put malware on major websites. Other stories cover AI-assisted exploitation, fake-session and phishing campaigns aimed at IT staff and gamers, and data breaches that bring legal fallout. For defenders, the common lesson is that patch speed, third-party script trust, and credential hygiene matter more than ever.
- Google reported that an AI agent turned a published security patch into a working exploit with no human guidance, showing how quickly patch diffs can be weaponized.
- Akamai published a blueprint aimed at helping federal agencies meet the requirements of CISA's Binding Operational Directive 26-04.
- A startup founded by three former Freshworks executives raised $53 million to build an AI-native CRM to compete with Salesforce and HubSpot.
- Attackers hijacked the Optimizely experimentation platform to deliver malware through 178 popular websites, showing how third-party scripts can become a single point of failure.
- A fake Prime Gaming reward page for EA Sports FC 27 is tricking players into entering their EA passwords and two-factor codes, which attackers can then use to take over accounts.
- A ClickFix social-engineering campaign of unknown origin is delivering the NetSupport RAT, giving attackers remote control of infected machines.
- Three proposed class actions filed by five alumni accuse Boston University of failing to protect personal data stored in its student information system.
- The FBI warns that criminals are abusing the ARTEX AI security-testing agent, with some intrusions reaching privileged access within three hours.
- Huntress published a retrospective on its CTF 2026 event, which tested how AI tools can support human problem-solving in realistic security challenges.
- A Check Point article argues that vulnerability management does not end at discovery, since assessment, ownership, prioritization, and verification often determine how long a flaw stays open.
- A critical flaw dubbed Java4Shell (CVE-2026-114938) in OpenJDK's Base64 decoder allows unauthenticated remote code execution, and Oracle released fixes on November 17.
- Huntress redesigned its former Threat Report as a one-page Security Insights Report in an updated reporting center.
- Attackers are exploiting CVE-2026-42271 and CVE-2026-48710 in LiteLLM and Starlette, which can be chained for unauthenticated remote code execution.
- Old Bitly short links in social posts from brands such as McDonald's and Burger King are being hijacked to redirect visitors to unrelated Wikipedia pages.
- Malware was silently injected into major sites including the BBC, CNN, Equifax, Ledger, and Zoom via the Optimizely compromise, exploiting weak controls over third-party content.
- Researchers demonstrated remote code execution in Telegram's newly launched widgets, and Kaspersky reports suspected active exploitation to deliver WebRAT malware.
- Huntress found more than 2,000 fake guest sessions on self-hosted ScreenConnect servers, disguised as cryptocurrency exchange machines and targeting IT technicians at 109 organizations.
- Forescout reported that AI cut vulnerability discovery time from nine months to 90 hours in a new study of AI-assisted security research.
- A guide explains how to prepare digital accounts for death so that loved ones can manage online accounts and preserve important files.
- An essay argues that AI speeds up software creation more than it removes the work of operating infrastructure, so Kubernetes remains relevant, citing a Zapier survey in which only 30% of agent-using leaders said agents completed over 60% of tasks unassisted.
- The NSANet classified internal network at the National Security Agency went down, according to a defense official, disrupting agency operations.
- Alibaba released Qwen4-Max, claiming a 62% lower average task cost than its predecessor, and added safeguards against offensive cyber use, though analysts caution the figure is a vendor benchmark.
- Cisco Talos is tracking CVE-2026-114938 ("Java4Shell") with a CVSS 4.0 score of 9.2, warning it may enable unauthenticated remote code execution.
- The SANS Internet Storm Center's daily Stormcast podcast for November 18, 2026 covered current threat activity.
- A separate blog post describes CVE-2026-114938 as a critical memory corruption vulnerability in OpenJDK Java, reinforcing the urgency of patching.
Summaries written with AI from the linked source articles.
Sources in this roundup
| CyberMaterial |
|
5 article(s) |
| Malwarebytes |
|
4 article(s) |
| Blog |
|
2 article(s) |
| Huntress Blog |
|
2 article(s) |
| InfoWorld |
|
2 article(s) |
| Check Point Blog |
|
1 article(s) |
| Cisco Talos Blog |
|
1 article(s) |
| Cyber Security News |
|
1 article(s) |
| IT SECURITY GURU |
|
1 article(s) |
| Krebs on Security |
|
1 article(s) |
| Malware-Traffic-Analysis.net – Blog Entries |
|
1 article(s) |
| Risky Bulletin |
|
1 article(s) |
| SANS Internet Storm Center, InfoCON: green |
|
1 article(s) |
| securityweek |
|
1 article(s) |
| www.theregister.com – Articles |
|
1 article(s) |
Most-mentioned keywords
| security |
|
4 mention(s) |
| vulnerability |
|
4 mention(s) |
| critical |
|
3 mention(s) |
| cve |
|
3 mention(s) |
| java |
|
3 mention(s) |
| openjdk |
|
3 mention(s) |
| agents |
|
2 mention(s) |
| discovery |
|
2 mention(s) |
Sources
- Google: AI agent turned a security patch into a working exploit with no human help
- CISA BOD 26-04: Akamai’s Strategic Blueprint for Federal Security
- Freshworks alums raise $53M for AI-native CRM
- Risky Bulletin: Optimizely hijacked to deliver malware
- Fake Prime Gaming x FC 27 reward site harvests EA passwords and 2FA codes
- 2026-11-16: Unidentified ClickFix pushes NetSupport RAT
- Boston U. alumni sue over data breach
- Criminals are using ARTEX AI agents to speed up cyberattacks, FBI warns
- The AI Arena: Huntress CTF 2026 Retro
- The Vulnerability Management Lifecycle: What Happens After Discovery?
- Critical Java4Shell Flaw Lets Hackers Execute Code Without Authentication on OpenJDK Servers
- New Huntress Security Insights Report & Reporting Center
- LiteLLM/Starlette CVE Exploitation in Wild
- Hijacked Bitly links turn old brand posts into offensive redirects
- Attackers Used Optimizely to A/B Test Malware on Top Websites
- Telegram Widgets Exploited for RCE and WebRAT
- Fake “Binance” and “KuCoin” machines planted on ScreenConnect servers in apparent trap for IT technicians
- AI Accelerates Vulnerability Discovery to 90 Hours
- How to prepare your digital accounts for after your death
- Will agents make Kubernetes obsolete?
- Classified Communications Network at the National Security Agency Goes Down, Defense Official Says
- Alibaba touts 62% lower task costs with Qwen4-Max, adds cyber safeguards
- CVE-2026-114938: Critical vulnerability in OpenJDK JAVA
- ISC Stormcast For Wednesday, November 18th, 2026 https://isc.sans.edu/podcastdetail/10184, (Wed, Nov 18th)
- CVE-2026-114938: Critical Memory Corruption Vulnerability in OpenJDK Java
