CVE-2026-65315: Ollama (HEAD f0078ae) contains an uncontrolled memory allocation vulnerability in the GGUF metadata parser that allows r

Ollama (a tool for running AI models locally) has a vulnerability in its GGUF metadata parser (the code that reads model file headers) that allows attackers to crash the server by uploading a specially crafted model file with fake size information. The parser doesn't check if the claimed sizes match the actual file, so it tries to allocate huge amounts of memory and crashes the entire server.