CVE-2025-53630

CVE-2025-53630 affects llama.cpp (ggml/gguf.cpp) where an integer overflow in gguf_init_from_file_impl can cause a heap out-of-bounds read/write. The vulnerability impacts inference paths in llama.cpp and is fixed by commit 26a48ad699d50b6268900062661bd22f3e792579. Connected sources document the ...