Advisories
Security vulnerabilities, privacy incidents, safety concerns, and policy updates affecting LLMs and AI agents.
Security vulnerabilities, privacy incidents, safety concerns, and policy updates affecting LLMs and AI agents.
3,134 items
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a path traversal vulnerability, tracked as CVE-2026-97671. A remote authenticated attacker can exploit it to obtain sensitive information.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-97655. An incomplete blocklist in the code security scanner could allow a remote attacker to execute arbitrary code.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw in ZIP file extraction. A remote authenticated attacker can trigger uncontrolled resource consumption, causing a denial of service.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-93678. A remote authenticated attacker can obtain sensitive information because of improper authorization.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw that exposes sensitive information to an unauthorized actor. A remote attacker who has valid authentication can obtain that information.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-93675. A remote attacker could execute arbitrary code through an expected dependency confusion flaw.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw, tracked as CVE-2026-93674, caused by improper neutralization of special elements used in an OS command. A remote attacker could exploit it to execute arbitrary code.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-93449. A remote authenticated attacker can exploit improper control of code generation to execute arbitrary code.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw where a pathname is not properly limited to a restricted directory. A remote authenticated attacker can exploit this to obtain sensitive information.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-93447. An attacker who holds the server secret and has Redis write access can submit a malicious serialized cache value. When that value is retrieved, deserialization could execute attacker-controlled code with the privileges of the service process.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-93445. A remote authenticated attacker could exploit improper control of code generation to execute arbitrary code.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain CVE-2026-93443. A remote authenticated attacker can exploit improper neutralization of special elements used in code to execute arbitrary code.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-88962. A remote attacker with valid authentication can execute arbitrary code because the product improperly controls code generation.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw, tracked as CVE-2026-104334, caused by improper control of code generation. A remote attacker could exploit it to execute arbitrary code.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw in which a pathname is not properly limited to a restricted directory. A remote authenticated attacker can exploit this to obtain sensitive information.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-101331. A remote authenticated attacker can obtain sensitive information because credentials are insufficiently protected.
IBM Langflow OSS versions 1.0.0 through 1.12.2 are affected by CVE-2026-101329. A remote authenticated attacker can obtain sensitive information because of improper access control.
IBM Langflow OSS versions 1.0.0 through 1.12.2 contain a flaw tracked as CVE-2026-104335. Due to improper access control, a remote authenticated attacker could execute arbitrary code.
CVE-2026-105192, rated CVSS 9.8 critical, affects lmcache versions up to and including 0.3.9 in multiprocess (distributed) mode. The ZeroMQ ROUTER transport has no authentication, and a single crafted REGISTER_KV_CACHE frame reaches pickle.loads through DeviceIPCWrapper.Deserialize during argument decoding, executing commands as the LMCache process user, which is root in official container images. The score applies when the transport is bound to a routable address rather than the default localhost.
Fix: No fixed version has been published as of 2026-10-07. The source advises stopping network data from being passed to pickle, replacing the serializer behind msgpack extension code 1 with a safe format, and not calling pickle.loads on data that a
NVIDIA Model-Optimizer contains a deserialization of untrusted data vulnerability, tracked as CVE-2026-65142, that an attacker may trigger. A successful exploit might lead to code execution, data tampering, denial of service, and information disclosure.