Skip to content
Fencer Intel
medium 6.5

CVE-2026-85670 json

tokenizers BpeBuilder Buffer Overflow via merge token

published 2mo ago modified last mo. also EUVD-2026-71126

cvss
6.5
version 3.1, CNA
cisa kev
Not listed
not in CISA's catalog
epss
0.5%
EPSS 0.5% chance of exploitation in 30 days, percentile 40%
public exploits
0
none known

Summary

tokenizers (Hugging Face) is affected by an out-of-bounds buffer access in BpeBuilder::build (tokenizers/src/models/bpe/model.rs). When loading a tokenizer.json via Tokenizer::from_file/from_str, the builder sizes a scratch buffer to the longest vocabulary key, then writes each concatenated merge rule into it. A merge whose concatenated token exceeds the longest vocabulary key overruns the buffer, which Rust turns into a panic that aborts the process in Rust and FFI embeddings. This occurs at load time with no encoding required, so an attacker who supplies a crafted tokenizer.json can cause a denial of service. A secondary defect at the same location can cause a usize underflow (panic in debug, potential memory corruption in release) when continuing_subword_prefix is set and a merge token is shorter than the prefix. Observed in version 0.23.1.

Is CVE-2026-85670 in your stack? Fencer shows which of your repositories are affected as soon as an advisory lands, and opens the fix as a pull request.

Check with Fencer

Vendor advisories

What the vendors themselves say about their products: affected, fixed or not affected, and what to do.

red hat product security vex CVE-2026-85670 updated cvss 7.6

tokenizers: Hugging Face tokenizers: Denial of service and potential memory corruption via crafted file

  • Red Hat All currently supported Red Hat products
    not affected: vulnerable code not present

Red Hat CSAF VEX © Red Hat, Inc., from https://security.access.redhat.com/data/csaf/v2/vex/; changed: reduced to product statuses and remediations. CC BY 4.0 · csaf

CISA SSVC

Exploitation
Proof of concept is public
Automatable
no
Technical impact
partial
Assessed
2mo ago

Exploit prediction

0.47% chance of exploitation in the next 30 days, higher than 40% of all scored CVEs (FIRST EPSS, 2026-11-11).

Exploits and checks

  • github.com https://github.com/huggingface/tokenizers/issues/2094 referenced PoC, tagged "exploit" by a source, unverified

CVSS scores

assigned by version score vector
CNA 4.0 7.1 CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N
CNA 3.1 6.5 CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
CNA 4.0 7.1 CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:N/VI:N/VA:H/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

Inferred packages

No advisory lists a package for this vulnerability yet. The CVE names these as products and links their GitHub repository, and the package registry names the same repository, so they are probably the affected packages. Low confidence: the versions are the CVE’s own for the product, and a stack check reports the package as unchecked when there are none.

ecosystem package repository affected
npm tokenizers huggingface/tokenizers no versions given
pypi tokenizers huggingface/tokenizers no versions given

Affected products

  • huggingface tokenizers

Classification

Weakness

References

Sources

Fencer Intel keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.