Ollama Out-of-Bounds Read Vulnerability Allows Remote Process Memory Leak

Summary: Researchers disclosed CVE-2026-7482, a heap out-of-bounds read in Ollama's GGUF model loader before version 0.17.1, rated CVSS 9.1 and codenamed Bleeding Llama by Cyera, which the source says likely affects over 300,000 servers. An unauthenticated attacker can send a crafted GGUF file with declared tensor offset and size exceeding the file's actual length to the /api/create endpoint, causing the server to read past the allocated heap buffer during quantization and potentially leak process memory including environment variables, API keys, system prompts, and concurrent users' conversation data, which can be exfiltrated via /api/push to an attacker-controlled registry.
What's compelling: A single crafted model file can expose the API keys, system prompts, and live conversation data of every user on an exposed Ollama server, and the source notes the REST API ships without authentication out of the box.

Read the original at The Hacker News ↗

Topics: Vulnerabilities & Patching
Mentions: CVE-2026-7482
Related: ⚡ Weekly Recap: Linux Rootkit, macOS Crypto Stealer, Web Socket Skimmers and More Cyber Risks ·
Attackers Exploit UnrealIRCd Flaw That Enables Server-Level Command Injection Cyber Risks · Nov 14, 2026
U.S. CISA adds Check Point and UnrealIRCd flaws to its Known Exploited Vulnerabilities catalog Cyber Risks · Nov 14, 2026