{"action":"Upgrade llama.cpp to 5760+dfsg-1 or later.","affected":[{"fixed":"5760+dfsg-1","introduced":"0"}],"component":"llama.cpp","cve_ids":[],"ecosystem":"Debian","ecosystem_qualifier":"Debian:14","exploited":false,"fixed_versions":["5760+dfsg-1"],"id":"tcs_01M5NP0CB7RGSRCK7KQ5WVNKVG","modified_at":"2026-09-01T20:05:53.526948151Z","published_at":"2025-06-24T04:15:46.967Z","references":["https://osv.dev/vulnerability/DEBIAN-CVE-2025-52566","https://security-tracker.debian.org/tracker/CVE-2025-52566"],"severity":{"label":"high","score":8.8},"sources":[{"source":"osv","source_id":"DEBIAN-CVE-2025-52566","url":"https://osv.dev/vulnerability/DEBIAN-CVE-2025-52566"}],"summary":"llama.cpp is an inference of several LLM models in C/C++. Prior to version b5721, there is a signed vs. unsigned integer overflow in llama.cpp's tokenizer implementation (llama_vocab::tokenize) (src/l"}