Read the docs
TI-2026-58782 Aliases CVE-2026-85670GHSA-jrf5-f6jp-7vxc

tokenizers BpeBuilder Buffer Overflow via merge token

Priority MEDIUM 64CWE-787Published · Updated

Key facts

Attack vector
Network
Privileges required
None
User interaction
Passive
Attack complexity
Low
Attack requirements
None
Impact type
Denial of service
Confidentiality · integrity · availability
None · None · High
Fix status
No fix listed

Attack profile from CVSS 4.0 · CNA. Fix status from CNA. Impact type from CWE-787 and the CVSS impact.

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.

Why it matters now

Exploitation Proof of concept CISA SSVC · not in CISA KEV
EPSS 0.00467 40th percentileScored 2026-11-07
Fix No fix listed No fixed version in the sources yet

SSVC decision points

CISA ADP · SSVC 2.0.3

Exploitation
Poc
Automatable
No
Technical impact
Partial

The decision (Track, Attend, Act) also depends on your mission and well-being impact; see CISA's SSVC guide.

Affected products

VendorProductAffected versionsFixed inStatusSource
huggingfacetokenizersup to 0.23.2—No fix listedCNA
  • tokenizers Vendorhuggingface Affected up to 0.23.2 Fixed in — No fix listed CNA

Remediation

No fixed version listed by the sources yet.

Vendor assessments

Red Hat

Red Hat's page for this CVE →

Not affected: 1 product
  • Red Hat

Timeline

  1. 2026-10-02 Threatist record created from CVE-2026-85670
  2. 2026-10-03 Alias added: GHSA-jrf5-f6jp-7vxc
  3. 2026-11-06 Affected versions changed · Red Hat
  4. 2026-11-06 CVSS score added · Red Hat

References

Other 3

  • TI-2026-48351 sentence-transformers Arbitrary Code Execution on Local Model Load Despite trust_remote_code=False CVSS 4.0CRITICAL 9.3
  • TI-2026-20914 LeRobot Unsafe Deserialization Remote Code Execution via gRPC CVSS 4.0CRITICAL 9.3
  • TI-2026-04577 Policy Bypass in LightGlue Nested Config Resolution in huggingface/transformers CVSS 3.1CRITICAL 9.6
  • TI-2025-04589 Sandbox Escape Vulnerability in huggingface/smolagents CVSS 3.1CRITICAL 10.0
  • TI-2026-79938 Pickle deserialization in Hugging Face Datasets allows code execution CVSS 4.0HIGH 8.7
  • TI-2026-32710 Diffusers: None.py Trust Remote Code Bypass CVSS 3.1HIGH 8.8
See all 58 →

Sources

This product uses the NVD API but is not endorsed or certified by the NVD. Sources and licences