CVE-2026-85670: Memory Corruption (CVSS 6.5)

Description

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 your site exposed to CVE-2026-85670?

Run a free security scan — no signup, results in seconds.

CVSS v3.1 Score

6.5

MEDIUM

CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H

EPSS — Exploit Prediction

0.0031

Probability of exploitation

0.24%

Percentile rank

EPSS estimates the probability that this vulnerability will be exploited in the wild within the next 30 days. A higher score means more likely to be exploited.

Weakness Type (CWE)

CWE-787 Out-of-bounds Write

References

Other References

https://github.com/huggingface/tokenizers https://github.com/huggingface/tokenizers/blob/v0.23.2/tokenizers/src/models/bpe/model.rs https://github.com/huggingface/tokenizers/issues/2094 https://www.vulncheck.com/advisories/tokenizers-bpebuilder-buffer-overflow-via-merge-token https://github.com/huggingface/tokenizers/issues/2094

Frequently Asked Questions

What is CVE-2026-85670? +

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. It has a CVSS v3.1 base score of 6.5 (MEDIUM).

How severe is CVE-2026-85670? +

CVE-2026-85670 has a CVSS v3.1 score of 6.5 out of 10, rated MEDIUM. This is a medium-severity vulnerability that should be remediated as part of regular maintenance. The EPSS score is 0.0031, placing it in the 0th percentile for exploitation probability.

How do I check if I'm vulnerable to CVE-2026-85670? +

You can use Secably's free Website Scanner to check your website for known vulnerabilities. For infrastructure scanning, use the Port Scanner to identify exposed services that may be affected. Check the vendor advisories linked above for specific patch and version information.

Related Vulnerabilities

CVE-2026-8358

LibreOffice Calc can import tracked changes from a spreadsheet document. A heap buffer overflow existed when a document reused the …

CVE-2026-8356

LibreOffice can import presentations in the legacy binary PPT format. A stack buffer overflow existed when importing a colour-replacement record. …

CVE-2026-6040

A heap use-after-free existed when importing the blank-width characters of an ODF number format. A position value read from the …

CVE-2026-6039

LibreOffice can import drawings in the DXF format used by CAD software. A heap buffer overflow existed when importing a …

CVE-2026-3298

The method "sock_recvfrom_into()" of "asyncio.ProacterEventLoop" (Windows only) was missing a boundary check for the data buffer when using nbytes parameter. …

CVE-2026-14297

A buffer overflow in the Bluetooth Continuous Glucose Monitoring Service (CGMS) Record Access Control Point (RACP) write handler allows an …

CVE-2026-85670