On 5 August 2026, CISA added three actively exploited vulnerabilities to its Known Exploited Vulnerabilities catalog in a single batch: a critical remote-code-execution flaw in Langflow (CVE-2026-9198, CVSS 9.8), a cluster-encryption bypass in Apache Tomcat (CVE-2026-34486, CVSS 7.5), and an authentication-bypass flaw in N-able's N-central remote-management platform (CVE-2026-18556, plus CVE-2026-18577 covering an incomplete first fix — the same N-central bug this desk covered on 4 August). All three are confirmed under active attack, not just theoretically exploitable, and federal US agencies have until end of business Friday, 7 August, to apply mitigations.
Langflow is the most urgent of the three for anyone experimenting with AI tooling: the flaw lets an unauthenticated attacker run arbitrary code on a default install, and a fully working proof-of-concept has been circulating publicly since late July — weeks before CISA's listing caught up. The fix, version 1.10.1, shipped in July, but any internet-reachable instance still on an older build is an open door. The Tomcat flaw is older news to patch (fixed back in April, in versions 11.0.21, 10.1.54 and 9.0.117) but newer news to worry about: CISA attributes active exploitation of it to a Chinese-speaking group researchers track as "knaithe"/"KnYuan," who reportedly used the DeepSeek model through an offensive automation framework called Hermes Agent to probe more than 460 internet-exposed targets, mixing autonomous scanning with manual follow-up.
What this means for your business
If Apache Tomcat runs anywhere in your stack — a Java backend, a legacy internal tool, a client's custom application — check the exact version this week. Anything older than 11.0.21, 10.1.54 or 9.0.117 needs the April patch applied now rather than scheduled for later, because CISA is telling you the exploitation is live, not hypothetical. If you or a client has stood up Langflow to prototype an AI feature, don't treat it as a side project: an internet-reachable instance on 1.10.0 or earlier is a working remote-code-execution target with a public exploit already in circulation, and 1.10.1 closes it. And if you patched N-central after our 4 August story, go back and confirm you're on the complete fix (2026.3.1.7 or later, closing CVE-2026-18577) — CISA formally listing it as under active exploitation removes any "wait and see" option for anyone still running the incomplete patch. The pattern worth remembering across all three: Tomcat's bug was fixed in April and is still being exploited in August, purely because unpatched instances stayed online. Applying a fix once is not the same as confirming, months later, that every instance actually got it.