{"id":"CVE-2026-93348","aliases":["EUVD-2026-88318","GHSA-6crx-pj66-wf6h"],"status":"published","title":"Unsloth Zoo Code Injection via model_type in config.json","description":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the get_transformers_model_type() function in hf_utils.py collects model_type values from nested model configurations without enforcing a character allowlist, allowing newlines and arbitrary Python source to survive normalization. Attackers can embed a newline in a nested model_type value within a malicious model's config.json to terminate the generated import statement and execute arbitrary Python code via exec() in unsloth_compile_transformers(), achieving remote code execution as the loading user when the model is loaded for training or inference.","severity":{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-09","score":0.00409,"percentile":0.35451,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-94"],"published":"2026-09-28T15:09:41Z","modified":"2026-10-28T02:30:49Z","products":[{"vendor":"unslothai","product":"unsloth"},{"vendor":"unslothai","product":"unsloth-zoo"}],"references":[{"url":"https://github.com/unslothai/unsloth-zoo/pull/1083","tags":["patch","issue-tracking","fix"]},{"url":"https://github.com/unslothai/unsloth-zoo/pull/1108","tags":["patch","issue-tracking","fix"]},{"url":"https://github.com/unslothai/unsloth/commit/92ee0207a38ebc8af1a0e678aa92d7e7d901ea0d","tags":["patch","fix"]},{"url":"https://www.vulncheck.com/advisories/unsloth-zoo-code-injection-via-model-type-in-config-json","tags":["third-party-advisory","advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-93348","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-6crx-pj66-wf6h"},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/93xxx/CVE-2026-93348.json","tags":["advisory"]},{"url":"https://github.com/unslothai/unsloth","tags":["package"]},{"url":"https://github.com/unslothai/unsloth-zoo","tags":["package"]}],"sources":{"cve":[{"id":"CVE-2026-93348","title":"Unsloth Zoo Code Injection via model_type in config.json","description":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the get_transformers_model_type() function in hf_utils.py collects model_type values from nested model configurations without enforcing a character allowlist, allowing newlines and arbitrary Python source to survive normalization. Attackers can embed a newline in a nested model_type value within a malicious model's config.json to terminate the generated import statement and execute arbitrary Python code via exec() in unsloth_compile_transformers(), achieving remote code execution as the loading user when the model is loaded for training or inference.","status":"published","published":"2026-09-28T15:09:41.524Z","modified":"2026-09-30T14:43:10.011Z","cvss":[{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high","assessor":"cna"},{"version":"3.1","score":8.1,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:N","level":"high","assessor":"cna"}],"cwes":["CWE-94"],"updated_at":"2026-11-02T14:19:54.122945Z"}],"euvd":[{"id":"EUVD-2026-88318","description":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the get_transformers_model_type() function in hf_utils.py collects model_type values from nested model configurations without enforcing a character allowlist, allowing newlines and arbitrary Python source to survive normalization. Attackers can embed a newline in a nested model_type value within a malicious model's config.json to terminate the generated import statement and execute arbitrary Python code via exec() in unsloth_compile_transformers(), achieving remote code execution as the loading user when the model is loaded for training or inference.","status":"published","published":"2026-09-28T15:09:41Z","modified":"2026-09-30T14:43:10Z","cvss":[{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high"}],"updated_at":"2026-11-02T17:27:49.499669Z"}],"ghsa":[{"id":"GHSA-6crx-pj66-wf6h","title":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8...","description":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the get_transformers_model_type() function in hf_utils.py collects model_type values from nested model configurations without enforcing a character allowlist, allowing newlines and arbitrary Python source to survive normalization. Attackers can embed a newline in a nested model_type value within a malicious model's config.json to terminate the generated import statement and execute arbitrary Python code via exec() in unsloth_compile_transformers(), achieving remote code execution as the loading user when the model is loaded for training or inference.","status":"published","published":"2026-09-28T18:31:24Z","modified":"2026-09-28T18:31:33Z","cvss":[{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high"},{"version":"3.1","score":8.1,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:N","level":"high"}],"cwes":["CWE-94"],"updated_at":"2026-11-02T15:12:32.639211Z"}],"nvd":[{"id":"CVE-2026-93348","description":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the get_transformers_model_type() function in hf_utils.py collects model_type values from nested model configurations without enforcing a character allowlist, allowing newlines and arbitrary Python source to survive normalization. Attackers can embed a newline in a nested model_type value within a malicious model's config.json to terminate the generated import statement and execute arbitrary Python code via exec() in unsloth_compile_transformers(), achieving remote code execution as the loading user when the model is loaded for training or inference.","status":"published","published":"2026-09-28T16:17:17.643Z","modified":"2026-09-30T17:32:07.107Z","cvss":[{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high","assessor":"secondary"},{"version":"3.1","score":8.1,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:N","level":"high","assessor":"secondary"}],"cwes":["CWE-94"],"updated_at":"2026-11-02T14:12:52.081323Z"}],"osv":[{"id":"CVE-2026-93348","title":"Unsloth Zoo Code Injection via model_type in config.json","description":"Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the get_transformers_model_type() function in hf_utils.py collects model_type values from nested model configurations without enforcing a character allowlist, allowing newlines and arbitrary Python source to survive normalization. Attackers can embed a newline in a nested model_type value within a malicious model's config.json to terminate the generated import statement and execute arbitrary Python code via exec() in unsloth_compile_transformers(), achieving remote code execution as the loading user when the model is loaded for training or inference.","status":"published","published":"2026-09-28T15:09:41.524Z","modified":"2026-10-28T02:30:49.314240556Z","cvss":[{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high"}],"cwes":["CWE-94"],"updated_at":"2026-11-02T15:45:20.931171Z"}]}}
