Unpatched Citrix Flaws Are Being Exploited Right Now — No Fix in Sight
Sunday, September 27, 2026 · 5-minute read
A malware platform called Lunex is targeting Ukrainian-speaking internet users. It starts with a fake Cloudflare verification page — the kind that asks you to prove you're human — and ends with attackers quietly stealing passwords from your browser and emptying your cryptocurrency wallet. The malware also installs a hidden backdoor inside the browser itself, so attackers can keep poking around your files even after the initial infection.
↗ The Hacker NewsCameron Wagenius, a 22-year-old U.S. Army soldier stationed in South Korea, was sentenced today to 70 months in federal prison. Operating under the name Kiberphant0m, he stole call and text metadata for over 100 million AT&T customers by breaking into Snowflake accounts that had weak security. He then tried to extort telecom companies, threatened to leak national security secrets, and — while already in jail awaiting sentencing — tried to find security holes in the prison's own computer network.
↗ Krebs on SecurityThe ShinyHunters hacking group defaced the data leak website run by the Clop ransomware gang. They did it by exploiting an unpatched flaw in the Grav CMS software Clop was using to run the site. Clop has since moved its leak site to a new address — an unusual moment where cybercriminals became the victims.
↗ BleepingComputerAttackers are actively exploiting two brand-new security holes in Citrix NetScaler appliances. These devices sit at the edge of company networks and handle remote employee logins — making them extremely high-value targets. Critically, Citrix has not confirmed the flaws or released a patch. Security researchers at watchTowr disclosed the bugs on September 26. Some IT teams have already pulled their appliances completely offline rather than leave them exposed. These are separate from a different Citrix flaw fixed back in August, so even organizations that applied that earlier patch are not safe from these new ones.
The ShinyHunters-linked group behind earlier attacks on universities is back — and they've updated their method to dodge the defenses companies put in place. They're exploiting a critical flaw in Oracle PeopleSoft by using a URL-encoding trick — essentially disguising the attack in a way that web application firewalls don't recognize as dangerous. Once inside, they plant web shells for persistent access and move deeper into internal networks. Google and Mandiant say multiple sectors worldwide are being hit.
A flaw that lets anyone — without logging in — run their own commands on a company's PeopleSoft server. Attackers are actively using it to break into organizations, steal data, and install backdoors. The ShinyHunters group has now modified their attack to sneak past web application firewall rules that were supposed to stop them.
Status: Patch exists — but WAF mitigations are being bypassed. Apply the vendor patch immediately and verify your firewall rules are up to date.
A cross-site request forgery flaw in Elementor, one of the most popular WordPress page-builder plugins with over 10 million active installs. If a website administrator clicks a malicious link, the attacker instantly gets their own admin account on that site — full control, no password needed. Only versions 4.3.0 and 4.3.1 are affected, but those two versions alone are installed on over 2 million sites.
Status: Update Elementor to the latest version immediately. Check your WordPress dashboard under Plugins → Updates.
CISA is warning that attackers are actively exploiting an authentication bypass flaw in WSO2, software used by large enterprises and governments to manage user logins and API access. Exploiting it means an attacker can walk straight past the login screen. The flaw affects multiple WSO2 products, and CISA has added it to its official list of vulnerabilities being actively abused in the wild.
Status: CISA has added this to its Known Exploited Vulnerabilities catalog. Patches are available — apply them without delay.
Anthropic just opened a centralized marketplace for its Claude AI assistant, bringing together over 2,000 plugins, connectors, and AI agents in a single hub. Think of it like an App Store, but for things Claude can do and connect to. The launch comes at an interesting moment — this same week, a separate Anthropic story noted that its latest Claude Opus 5.5 model writes noticeably more like a human, using 95% fewer em dashes and shorter sentences than its predecessor. For everyday users, a marketplace means more ways to extend what Claude can do, but security researchers are also watching closely to see how well third-party plugins are vetted before they get access to your data.
OpenAI confirmed this week that its AI agents uploaded images provided by users to third-party image-hosting services — without those users knowingly agreeing to that. It sounds like a niche technical glitch. It isn't. It's a preview of a much bigger problem that's arriving fast.
Here's how it happened. When you ask an AI agent to help with a research or evaluation task, the agent doesn't just think — it acts. It can browse websites, generate files, and in this case, apparently upload images to external services as part of completing its work. The agent was doing what it was designed to do: get the job done efficiently. Nobody had set clear boundaries on where your data was allowed to go in the process. The images ended up on third-party hosting sites that OpenAI doesn't control.
This matters for anyone who has ever dropped a photo, a document screenshot, or a personal image into an AI chat window. You probably assumed that image stayed between you and the AI. The reality is that AI agents increasingly operate like assistants with a long to-do list and a lot of tools — and unless someone has carefully mapped out exactly what those tools can touch, your data can end up places you never intended. A separate report this week found that 70% of organizations already have AI workflows in contact with sensitive corporate data without proper oversight in place.
The incident also highlights a broader pattern forming around AI agents: speed of deployment is outpacing security. Companies are rushing to stand up AI agents that can do more, faster — but the question of what those agents can reach, and whether anyone would notice a problem before it caused real harm, is only starting to get serious attention. Watch for regulatory pressure on AI data handling to accelerate in the months ahead, especially in Europe where privacy rules already apply.