CYBERSECURITYTRACKER
Permanent story citation

High-severity Nvidia bug could crash GPU monitoring on exposed servers

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 9806

As cited

Copy frozen at (site build).

vulnerabilities

High-severity Nvidia bug could crash GPU monitoring on exposed servers

Researchers discovered thousands of Nvidia GPU servers exposing DCGM Exporter, a monitoring service, to the internet without authentication. The exposure revealed detailed telemetry and hardware information for approximately 12,000 GPUs worth roughly $100 million across 300 organizations, with CVE-2026-47483 (CVSS 8.2) allowing unauthenticated attackers to crash the service through resource exhaustion. Nvidia released a fix in version 4.8.2 in September, and operators should restrict these monitoring services to authorized infrastructure only.

Why it matters: GPU cluster operators and data center managers need to immediately audit and restrict internet-facing DCGM Exporter and Prometheus instances, as exposed systems provide attackers reconnaissance data for targeting artificial intelligence (AI) infrastructure and can be disrupted via denial of service.

Source published
First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

vulnerabilities

High-severity Nvidia bug could crash GPU monitoring on exposed servers

Researchers discovered thousands of Nvidia GPU servers exposing DCGM Exporter, a monitoring service, to the internet without authentication. The exposure revealed detailed telemetry and hardware information for approximately 12,000 GPUs worth roughly $100 million across 300 organizations, with CVE-2026-47483 (CVSS 8.2) allowing unauthenticated attackers to crash the service through resource exhaustion. Nvidia released a fix in version 4.8.2 in September, and operators should restrict these monitoring services to authorized infrastructure only.

Why it matters: GPU cluster operators and data center managers need to immediately audit and restrict internet-facing DCGM Exporter and Prometheus instances, as exposed systems provide attackers reconnaissance data for targeting artificial intelligence (AI) infrastructure and can be disrupted via denial of service.

Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary