Ollama flaw could leak memory and expose secrets via crafted GGUF files
TL;DR Summary
A high-severity heap out-of-bounds read in Ollama's GGUF model loader (CVE-2026-7482, CVSS 9.1) lets remote, unauthenticated attackers leak the Ollama process memory by sending a crafted GGUF to /api/create, potentially exposing environment variables, API keys, prompts, and user data across 300k+ servers. Mitigations include applying fixes, restricting network exposure, and deploying a proxy/auth gateway. The article also cites two unpatched Windows updater flaws (CVE-2026-42248/42249) that can enable persistent code execution; recommended workarounds are to disable automatic updates and remove Startup-folder shortcuts until patches land.
- Ollama Out-of-Bounds Read Vulnerability Allows Remote Process Memory Leak The Hacker News
- Unpatched flaws turn Ollama’s auto-updater into a persistent RCE vector, researchers say Help Net Security
- Ollama vulnerability highlights danger of AI frameworks with unrestricted access csoonline.com
- Critical Bug Could Expose 300,000 Ollama Deployments to Information Theft SecurityWeek
- Major AI platform Ollama critically leaking: 300,000 servers exposed to hackers Cybernews
Reading Insights
Total Reads
0
Unique Readers
18
Time Saved
5 min
vs 6 min read
Condensed
92%
1,085 → 84 words
Want the full story? Read the original article
Read on The Hacker News