CISA Adds Three Actively Exploited Bugs to Its Must-Patch List
A critical Langflow flaw joins Apache Tomcat and N-central issues on the U.S. government's Known Exploited Vulnerabilities catalog after evidence of live attacks.
Key points
- CISA added three actively exploited flaws to its Known Exploited Vulnerabilities catalog on August 5, 2026.
- The headline bug is CVE-2026-9198, a critical code injection flaw in Langflow scoring 9.8 out of 10.
- The Langflow flaw lets an unauthenticated attacker run any code they like on a vulnerable server.
- U.S. federal agencies now have a hard deadline to patch or stop using the affected software.
The U.S. Cybersecurity and Infrastructure Security Agency, the federal body that helps protect American networks, has told government agencies to urgently fix three software bugs that criminals are already using in real attacks.
The agency, known as CISA, added the three flaws to its Known Exploited Vulnerabilities catalog on August 5, 2026. That catalog, often shortened to KEV, is basically a public list of bugs that are not theoretical: someone is actively abusing them right now. Federal agencies are legally required to patch anything on it fast.
Private companies are not bound by that rule. In practice, though, most security teams treat the KEV list as their real to-do list, because the bugs on it are the ones being weaponised in the wild.
What are the three flaws?
The standout is a critical bug in Langflow, an open-source tool that lets developers wire together AI models into working applications. Tracked as CVE-2026-9198, it scores 9.8 out of 10 on the industry severity scale. Anyone on the internet can send a crafted request and run their own code on the server, no password required. That is about as bad as it gets.
The other two flaws, first reported by The Hacker News, sit in Apache Tomcat, a widely used web server that runs behind countless corporate websites, and in N-central, a remote management tool that IT providers use to look after their customers' computers. The failure mode here is familiar: both products sit deep inside company networks, so a single working exploit can hand attackers the keys to a lot of doors at once.
| Product | CVE | Severity | Type of flaw |
|---|---|---|---|
| Langflow | CVE-2026-9198 | 9.8 (Critical) | Remote code execution |
| Apache Tomcat | Listed on KEV | Not disclosed | Actively exploited |
| N-central | Listed on KEV | Not disclosed | Actively exploited |
Why does this matter for ordinary people?
Most readers will never touch Langflow or Tomcat directly. But the banks, hospitals, retailers, and government services you rely on probably run one of them somewhere in their stack. When a management tool like N-central is broken into, the fallout usually lands on the small businesses whose IT is outsourced: the dentist's office, the local law firm, the corner accountancy.
One thing the post-mortem will say, as always, is that the patch was available and nobody applied it in time.
What should companies do now?
Treat all three as drop-everything work. If you run Langflow, upgrade to the fixed release and, more importantly, ask why an AI development tool was ever reachable from the public internet in the first place. That is a configuration question, not a patching one.
For Tomcat, check with your platform team which version is running and where. For N-central, follow the vendor's advisory and rotate any credentials the tool touches, because a break-in there rarely stays contained to the tool itself.
CISA's deadline for federal agencies is typically 21 days. Everyone else should aim for faster.
Operational takeaway: if your asset inventory cannot answer "do we run Langflow, and is it exposed?" in under an hour, that is the bug worth fixing first.
