Key Takeaways
- CISA has issued a three‑day deadline for U.S. federal agencies to remediate actively exploited vulnerabilities in IBM Langflow, N‑able N‑central, and Apache Tomcat.
- The most severe flaw is CVE‑2026‑9198 in IBM Langflow (CVSS 9.8), which lets an unauthenticated attacker chain two API endpoints to bypass authentication and execute arbitrary code.
- Proof‑of‑concept exploits for CVE‑2026‑9198 appeared publicly in late July, and a separate critical Langflow flaw (CVE‑2026‑0770) was already being abused for root‑level remote code execution.
- N‑central’s vulnerability CVE‑2026‑18576 permits admin‑account hijacking without authentication; an initial patch proved insufficient, prompting an emergency hotfix released in early August.
- Apache Tomcat’s CVE‑2026‑34486 (CVSS 7.5) stems from an incomplete fix for the earlier critical CVE‑2026‑29146; threat actors have used it to plant reverse shells on Tomcat servers.
- All three flaws have been added to CISA’s Known Exploited Vulnerabilities (KEV) catalog, though the agency has not disclosed whether they are employed in ransomware campaigns.
- Federal agencies must apply vendor‑provided mitigations or work‑arounds by the end of Friday, July 7, to comply with CISA’s directive.
- Organizations outside the federal sector should also prioritize patching, network segmentation, and monitoring for anomalous API or administrative activity to reduce risk.
Overview of CISA Directive and Timeline
The Cybersecurity and Infrastructure Security Agency (CISA) announced on July 4 that federal agencies have three days to mitigate three actively exploited vulnerabilities affecting IBM Langflow, N‑able N‑central, and Apache Tomcat. The directive, communicated via an urgent bulletin, requires agencies to apply available patches, configuration changes, or vendor‑recommended work‑arounds by the close of business on Friday, July 7. CISA emphasized that the vulnerabilities are being leveraged in the wild and have been added to its Known Exploited Vulnerabilities (KEV) catalog, underscoring the need for immediate remedial action across federal information systems.
IBM Langflow Critical Flaw – CVE‑2026‑9198
IBM’s Langflow, a visual framework for constructing AI agents, contains a critical vulnerability tracked as CVE‑2026‑9198 with a CVSS score of 9.8. The flaw resides in the default deployment’s RESTful API endpoints that, when chained together, allow an unauthenticated remote attacker to bypass authentication mechanisms and execute arbitrary code on the host. Because the attack requires no prior credentials, exposed Langflow instances are especially susceptible to automated scanning and exploitation by botnets.
Emergence of Public Proof‑of‑Concept Exploits
In late July, security researchers published multiple fully functional proof‑of‑concept (PoC) exploits for CVE‑2026‑9198, accompanied by step‑by‑step instructions on how to chain the vulnerable APIs. The rapid public release of these PoCs lowered the barrier for threat actors, leading to increased scanning activity against Langflow‑enabled environments. Security teams reported a surge in attempted exploitation attempts within days of the PoC disclosure, prompting CISA to issue an urgent advisory.
Related Langflow Vulnerability – CVE‑2026‑0770
Prior to the discovery of CVE‑2026‑9198, CISA had already issued an alert for another critical Langflow flaw, CVE‑2026‑0770, which also enables remote code execution with root privileges. Although the attack vector differs—leveraging a deserialization weakness in a plugin component—the outcome is similar: full control of the underlying server. The concurrent presence of two high‑impact Langflow vulnerabilities has amplified the urgency for organizations to either disable the framework pending patches or enforce strict network isolation.
N‑central Vulnerability – CVE‑2026‑18576
N‑able’s remote monitoring and management (RMM) platform, N‑central, contains a high‑severity vulnerability identified as CVE‑2026‑18576 (CVSS ≈ 8.1). The flaw permits an unauthenticated attacker to hijack administrative accounts by manipulating a token validation routine, effectively granting the attacker privileged access to managed endpoints. N‑able initially released a patch addressing the original issue, but subsequent analysis revealed that the fix did not fully close the exploitation window.
Patch Insufficiency and Emergency Hotfix
Threat actors quickly adapted to the incomplete patch, discovering a new bypass technique that retained the original vulnerability’s impact. N‑able warned customers on August 1 that active exploitation of the newly identified variant was underway, assigning it the same CVE identifier, CVE‑2026‑18576. In response, the vendor released an emergency hotfix on the following Sunday, urging all N‑central installations prior to version 2026.3 to apply the update immediately. The hotfix re‑enforces token validation and adds additional logging to detect abuse attempts.
Apache Tomcat Vulnerability – CVE‑2026‑34486
Apache Tomcat, the widely used Java servlet container, is affected by CVE‑2026‑34486, which carries a CVSS score of 7.5. The vulnerability originates from an incomplete remediation of the earlier critical flaw CVE‑2026‑29146 (CVSS 9.8), which involved missing encryption of sensitive data in certain Tomcat connector configurations. Because the initial fix only addressed part of the data‑handling pathway, attackers can still intercept or manipulate unencrypted information under specific conditions.
Link to Prior Critical Flaw and Exploitation Attempt
Researchers at Palo Alto Networks Unit 42 observed on July 30 that a Chinese‑speaking threat actor attempted to exploit CVE‑2026‑34486 in a manual campaign targeting nine Apache Tomcat servers. The actor sought to plant reverse shells, thereby establishing persistent backdoors for later lateral movement or data exfiltration. Although the attempt was detected and blocked in the observed instances, the activity confirms that the vulnerability is being actively probed in the wild.
CISA KEV Catalog Addition and Unknown Attack Vectors
Following the confirmed exploitation of all three flaws, CISA added CVE‑2026‑9198, CVE‑2026‑18576, and CVE‑2026‑34486 to its Known Exploited Vulnerabilities (KEV) catalog. The agency noted that while the vulnerabilities are certainly being used in attacks, it has not yet determined whether they are employed in ransomware operations, espionage campaigns, or other malicious activities. This uncertainty reinforces the recommendation to treat each flaw as a high‑priority risk regardless of the observed motive.
Recommended Mitigations and Best Practices
Federal agencies should take the following steps to satisfy CISA’s directive:
- IBM Langflow: Apply the latest vendor patch that secures the API authentication flow; if patching is not immediately feasible, disable public access to the Langflow API endpoints and enforce strict network‑level segmentation.
- N‑central: Deploy the emergency hotfix released on August 1; verify that the version number reflects the fix (2026.3 or later). Additionally, enforce multi‑factor authentication for administrative accounts and monitor authentication logs for anomalous token usage.
- Apache Tomcat: Upgrade to a version that includes the complete fix for CVE‑2026‑29146 (and thus CVE‑2026‑34486). Where upgrading is not possible, enable TLS encryption for all connectors and configure the server to reject unencrypted traffic for sensitive applications.
Across all products, agencies should ensure that vulnerability scanning tools are updated to detect the specific CVEs, that intrusion detection systems are tuned to flag the associated exploit patterns, and that incident response playbooks include scenarios for remote code execution via these flaws.
Broader Implications for Software Supply Chain and Vulnerability Management
The simultaneous emergence of critical flaws in an AI development framework, an RMM platform, and a ubiquitous web server highlights the expanding attack surface presented by modern software supply chains. Organizations must adopt a proactive vulnerability management posture that includes:
- Continuous monitoring of vendor security advisories and rapid patch deployment cycles.
- Implementation of zero‑trust network principles to limit lateral movement even if a perimeter device is compromised.
- Regular penetration testing and red‑team exercises focused on newly disclosed vulnerabilities.
By treating each disclosed CVE as a potential foothold, enterprises can reduce the likelihood that threat actors achieve persistent access or data exfiltration.
Note on Ancillary Content
The original source also contained a promotional reference to a Picus whitepaper discussing breach and attack simulation testing for SIEM and EDR rules. While not directly related to the CISA advisory, the reference underscores the growing interest in validating defensive controls against the very types of exploits described above. Security teams may find such resources useful for assessing whether their detection capabilities can identify the exploitation techniques associated with CVE‑2026‑9198, CVE‑2026‑18576, and CVE‑2026‑34486.
Conclusion
CISA’s three‑day remediation order for IBM Langflow, N‑able N‑central, and Apache Tomcat underscores the severity and active exploitation of CVE‑2026‑9198, CVE‑2026‑18576, and CVE‑2026‑34486. Federal agencies—and indeed all organizations using these technologies—must prioritize patching, configuration hardening, and vigilant monitoring to mitigate the risk of remote code execution, privileged account takeover, and data exposure. Prompt action, informed by the detailed technical characteristics of each vulnerability, will be essential to safeguarding critical infrastructure against the current wave of threats.