CVE-2026-76841
RCE via Xinference loading untrusted Hugging Face models (trust_remote_code enabled).
Is CVE-2026-76841 being exploited?
Not confirmed. CVE-2026-76841 does not appear in CISA's Known Exploited Vulnerabilities catalog, which records only exploitation that has been observed and reported publicly. That is evidence of absence of a report, not evidence the flaw is unattacked. EPSS currently estimates a 0.71% probability of exploitation in the next 30 days.
How severe is CVE-2026-76841?
CVE-2026-76841 is rated High with a CVSS score of 8.7. Severity describes how bad exploitation would be, not how likely it is: pair it with exploitation evidence before deciding what to patch first.
Is there a patch for CVE-2026-76841?
Yes. A fix has been recorded for CVE-2026-76841. The vendor advisory is the authority on the exact fixed version — apply it from there rather than from a summary.
What does CVE-2026-76841 affect?
CVE-2026-76841 affects Xinference < 2.12.0, Model servers using Hugging Face tokenizers, Worker processes handling model loads. Confirm the exact affected versions against the vendor advisory before deciding you are exposed.
What should I do about CVE-2026-76841?
Upgrade to Xinference 2.12.0, set XINFERENCE_TRUST_REMOTE_CODE=false, restrict/validate model uploads and isolate workers.
Exploitation status reflects CISA's KEV catalog as we last synced it. Check the catalog directly.
Attacker-supplied model/tokenizer can execute arbitrary Python in worker processes, enabling remote code execution and privilege escalation.
Immediate action required
- affected< 2.12.0
As published in the CVE Program record. A version outside these ranges is not a statement that it is unaffected — vendors sometimes understate a range, and distribution-backported builds carry upstream numbers that do not reflect what was patched into them.