{"id":"CVE-2026-84286","published":"2026-10-12T17:17:35.873","lastModified":"2026-10-16T16:00:33.533","description":"ExLlamaV3’s EXL3 loader uses an untrusted, model-supplied index to select a CUDA kernel without first verifying that the index is within bounds. A malicious model can force the code to read and invoke an out‑of‑range function pointer, resulting in a reproducible crash. Under specific conditions involving certain host memory layouts, it may be possible for an attacker to exploit this vulnerability to achieve controlled indirect kernel dispatch during model load.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://github.com/turboderp-org/exllamav3/pull/310","tags":[]},{"url":"https://kb.cert.org/vuls/id/369611","tags":[]},{"url":"https://www.kb.cert.org/vuls/id/369611","tags":[]}],"exploitRefs":[{"url":"https://github.com/turboderp-org/exllamav3/pull/310","tags":[]}],"hasPoc":true,"ai":null}