KEV Alert: Langflow exposes tokens and remote execution; Apex One on-prem provides code distribution - urgent parking

The U.S. agency CISA has added two vulnerabilities exploited in practice to its Known Exploited Vulnerabilities (KEV) catalogue, a warning that requires federal entities to urgently correct the failures and that serves as an alarm signal for companies and security administrators around the world. This is about CVE-2025-34291(CVSS 9.4), a Langflow source validation failure that allows remote code execution and tokens and key exposure, and CVE-2026-34926(CVSS 6.7), a traversal vulnerability in the on-premises version of Trend Micro Apex One which, with prior administrative access to the server, allows to modify a critical table to inject code into the agents.

The case of Langflow is particularly concerned by the combination of defects that facilitated real attacks: according to public analysis, the vector combines excessively permissive CORS, lack of CSRF protection and an endpoint that by design accepts code execution. The result is not only the taking over of the vulnerable instance, but the possibility of exporting and abusing all the keys and tokens stored in that workspace, which can cause commitments chained in cloud and SaaS services. Security investigations reported exploitation by Iran-related actors in campaigns that use these failures to achieve initial access to target networks, raising the risk for organizations with automated integrations or Langflow-based developments.

In the case of Trend Micro Apex One, vulnerability requires a different context: it only affects on-premises and requires that the attacker already have administrative access to the server (obtained by other means). Even so, Trend Micro confirmed attempts at exploitation in real environments, which places the focus on the need not to minimize vulnerabilities that require "pre-conditions" because they are often part of attack chains. Modifying the server key table to distribute malicious code to agents can allow for persistence and lateral movement within business environments protected by that same EDR.

The inclusion in the KEV catalogue implies an operational obligation for federal civil agencies with a deadline on 4 June 2026 but the lesson for the private sector is clear: when there is evidence of active exploitation, the risk window closes very quickly. The immediate actions to be prioritized by the technical equipment are to apply the official patches as soon as they are available, to verify communications from the supplier and to follow the published guides; for Trend Micro it is appropriate to consult the manufacturer's alerts and recommendations at its support and advisory centre, and for the KEV entries the CISA itself keeps up-to-date information on mitigation and reference. More information and the full catalogue are available on the CISA website: Known Exploited Vulnerabilities Catalog and on the Trend Micro support portal: Trend Micro Support. For the context of CVSS vulnerabilities and scores, the NVD database may be useful: NVD.

Apply patches must be accompanied by forensic and containment measures: audit administrative access to the server, review login and authentication vectors, immediately rotate and revoke all keys, tokens and credentials stored in affected instances and restore agents from clean images when there is suspicion of manipulation. In cloud or SaaS environments where Langflow is integrated with multiple APIs, it is appropriate to carry out a comprehensive review of delegated permits and tokens, implement stricter CORS / CSRF rules where possible and place additional controls such as WAF and detect anomalies in endpoints calls that allow code execution.

For organizations that depend on on on-premises solutions like Apex One, priority should include checking server access settings, applying minimum privilege principles and network segmentation, and deploying specific monitoring to detect attempts to modify tables or distribute malicious packages to agents. If there is already evidence of commitment, it is reasonable to isolate the server concerned and consider a response plan that includes restoration from verified backup and, where appropriate, notification to the parties concerned and regulatory authorities.

Beyond the point patches, these incidents highlight two structural risks: the fragility introduced by tools that allow remote execution by design and the speed at which actors with capacity - such as groups attributed in public reports - make a vulnerability a gateway to critical infrastructure. The defence requires not only patching but also maintaining policies for the rotation of credentials, robust access controls, segmentation, and incident response exercises that include review of exposed integrations and secrets. The time between the publication of a vulnerability and its active exploitation may be short; the difference between effective mitigation and a larger incident is often dependent on the preparation and speed of response.