End of life: 11/4/2029, End of support: 11/4/2027, Latest version: 621.14

End of life: 11/4/2029, End of support: 11/4/2027, Latest version: 620.70.06

NVIDIA PhysX for all platforms contains a vulnerability where an attacker could cause an out-of-bounds write. A successful exploit of this vulnerability might lead to code execution, data tampering, denial of service, information disclosure.

NVIDIA NeMo Agent Toolkit contains a vulnerability where a attacker could cause a missing authentication issue. This vulnerability might lead to code execution, escalation of privileges, data tampering, denial of service, and information disclosure.

NVIDIA NeMo Agent Toolkit contains vulnerability where an attacker could cause a SQL injection issue. A successful exploit of this vulnerability might lead to data tampering, denial of service and information disclosure.

NVIDIA TensorRT-LLM contains a vulnerability where an attacker could cause a server-side request forgery. A successful exploit of this vulnerability might lead to information disclosure.

NVIDIA TensorRT-LLM contains a vulnerability where an attacker could cause a server-side request forgery. A successful exploit of this vulnerability might lead to information disclosure and data tampering.

NVIDIA TensorRT-LLM for any platform contains a vulnerability in KvCacheAwareRouter tokenizer loading, where an attacker could cause inclusion of functionality from an untrusted control sphere. A successful exploit might lead to code execution, information disclosure, data tampering, and denial of service.

This vulnerability allows remote attackers to execute arbitrary code on affected installations of NVIDIA Megatron Bridge. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The ZDI has assigned a CVSS rating of 7.8. The following CVEs are assigned: CVE-2026-24250.

This vulnerability allows remote attackers to execute arbitrary code on affected installations of NVIDIA Megatron Bridge. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The ZDI has assigned a CVSS rating of 7.8. The following CVEs are assigned: CVE-2026-24250.

NVIDIA Megatron-LM contains a vulnerability in checkpoint loading where an Attacker may cause an RCE by convincing a user to load a maliciously crafted file. A successful exploit of this vulnerability may lead to code execution, escalation of privileges, information disclosure, and data tampering.

NVIDIA Megatron-LM for all platforms contains a vulnerability in inference, where an attacker could cause hash collisions in the cache used for inference requests. A successful exploit of this vulnerability might lead to data tampering and partial information disclosure.

NVIDIA Megatron-LM contains a vulnerability in checkpoint loading where, an attacker could cause an RCE by convincing a user to load a maliciously crafted file. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, information disclosure, and data tampering.

NVIDIA NVSentinel for Linux contains a vulnerability in the platform-connectors component, where the gRPC health-event interface, exposed over a world-writable Unix domain socket, lacks authentication. A successful exploit of this vulnerability might lead to denial of service.

NVIDIA AIStore for Linux and Kubernetes contains a missing authorization vulnerability in the ETL management API where an unauthenticated remote attacker may access ETL information or control existing ETL jobs by sending crafted GET or POST requests to an AIStore proxy.

A successful exploit may lead to information disclosure and unauthorized start or stop operations that affect the integrity and availability of ETL data-processing pipelines.

NVIDIA AIStore for Linux contains a path traversal vulnerability in the dSort archive extraction component where a remote attacker may cause an improper limitation of a pathname to a restricted directory by supplying crafted archive member names to a dSort job. A successful exploit may lead to arbitrary file creation or overwrite, code execution, information disclosure, data tampering, and denial of service.

NVIDIA AIStore for Linux contains a missing authentication vulnerability in the token-revocation endpoint where an unauthenticated remote attacker may submit crafted token-revocation lists. A successful exploit may lead to data tampering and denial of service.

NVIDIA AIStore for Linux contains a vulnerability in the distributed sort component where an attacker could cause server-side request forgery by supplying a crafted external key map URL. A successful exploit of this vulnerability might lead to information disclosure.

NVIDIA AIStore for Linux contains a vulnerability in the experimental HT backend where an unauthenticated remote attacker may cause server-side request forgery by supplying a crafted original URL. A successful exploit of this vulnerability may lead to information disclosure.

NVIDIA AIStore for Linux contains a vulnerability in the legacy ETL Kubernetes PodSpec initialization API where a remote attacker may supply Kubernetes host-namespace and container-security settings. A successful exploit may lead to code execution, privilege escalation, information disclosure, data tampering, and denial of service.

NVIDIA AIStore for Linux contains a vulnerability in the Promote API where an unauthenticated remote attacker may read files accessible to the target process by supplying a crafted absolute source path. A successful exploit may lead to information disclosure.

NVIDIA AIStore for Linux contains a path traversal vulnerability in the target-side object rename handler, where an unauthenticated remote attacker could cause an improper limitation of a pathname to a restricted directory by submitting a crafted rename destination directly to an AIStore target, resulting in the arbitrary creation or overwrite of a file. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, information disclosure, data tampering, and denial of service.

NVIDIA Warp contains a vulnerability where an attacker could cause a heap-based buffer overflow. A successful exploit of this vulnerability might lead to denial of service.

NVIDIA DALI contains a vulnerability where an attacker could cause a stack-based buffer overflow. A successful exploit of this vulnerability might lead to code execution, information disclosure, data tampering, and denial of service.

NVIDIA DALI contains a vulnerability where an attacker could cause an uncontrolled out-of-bounds read. A successful exploit of this vulnerability mighty lead to information disclosure, data tampering, and denial of service.

NVIDIA DALI contains a vulnerability where an attacker could cause an uncontrolled out-of-bounds write. A successful exploit of this vulnerability might lead to denial of service, information disclosure, and data tampering.

NVIDIA DALI contains a vulnerability where an attacker could cause an uncontrolled out-of-bounds read. A successful exploit of this vulnerability might lead to information disclosure and denial of service.

NVIDIA Isaac Lab contains a vulnerability where an attacker could cause an eval injection. A successful exploit of this vulnerability might lead to code execution, information disclosure, data tampering, and denial of service.

TL;DR: NVIDIA DCGM Exporter is a widely used monitoring tool that collects GPU metrics like utilization, memory, temperature, and power consumption from AI servers. We discovered that thousands of these monitoring endpoints are exposed to the public internet, allowing anyone to inspect GPU infrastructure. Worse, we found a high-severity vulnerability (CVE-2026-47483, CVSS 8.2) that lets unauthenticated attackers exhaust server resources, crash GPU monitoring, and potentially disrupt AI workloads remotely. A few other interesting findings: 2,100+ exposed GPU servers, revealing telemetry from 12,000+ GPUs worth \~$100M, including H100s, H200s, and Blackwell GPUs. 60% of exposed GPUs reported 0% utilization across our scans. 12,096 additional servers exposed infrastructure details, including OS, firmware, and networking information. NVIDIA fixed the vulnerability following our responsible disclosure. Full research: https://lava.security/research/cve-2026-47483-nvidia-dcgm-exporter-vulnerability

NVIDIA Model-Optimizer contains a vulnerability where an attacker may cause deserialization of untrusted data. A successful exploit of this vulnerability might lead to code execution, data tampering, denial of service, and information disclosure.