CVE-2026-85670 (GCVE-0-2026-85670)
Vulnerability from cvelistv5 – Published: 2026-09-04 14:32 – Updated: 2026-09-04 17:49
VLAI
EPSS
VEX
Title
tokenizers BpeBuilder Buffer Overflow via merge token
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.
Severity
SSVC
Exploitation: poc
Automatable: no
Technical Impact: partial
CISA Coordinator · CISA-ADP (v2.0.3)
Decision recorded 2026-09-04 17:49 UTC
CWE
- CWE-787 - Out-of-bounds Write
Assigner
References
4 references
| URL | Tags |
|---|---|
| https://github.com/huggingface/tokenizers/issues/2094 | issue-tracking |
| https://github.com/huggingface/tokenizers | product |
| https://github.com/huggingface/tokenizers/blob/v0… | technical-description |
| https://www.vulncheck.com/advisories/tokenizers-b… | third-party-advisory |
Impacted products
1 product
| Vendor | Product | Version | CPE status | |
|---|---|---|---|---|
| huggingface | tokenizers |
Affected:
0 , ≤ 0.23.2
(semver)
|
guessed |
Date Public
2026-06-10 00:00
{
"containers": {
"adp": [
{
"metrics": [
{
"other": {
"content": {
"id": "CVE-2026-85670",
"options": [
{
"Exploitation": "poc"
},
{
"Automatable": "no"
},
{
"Technical Impact": "partial"
}
],
"role": "CISA Coordinator",
"timestamp": "2026-09-04T17:49:21.636135Z",
"version": "2.0.3"
},
"type": "ssvc"
}
}
],
"providerMetadata": {
"dateUpdated": "2026-09-04T17:49:42.263Z",
"orgId": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
"shortName": "CISA-ADP"
},
"references": [
{
"tags": [
"exploit"
],
"url": "https://github.com/huggingface/tokenizers/issues/2094"
}
],
"title": "CISA ADP Vulnrichment"
}
],
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"packageURL": "pkg:pypi/tokenizers",
"product": "tokenizers",
"vendor": "huggingface",
"versions": [
{
"lessThanOrEqual": "0.23.2",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "finder",
"value": "George Chen"
}
],
"datePublic": "2026-06-10T00:00:00.000Z",
"descriptions": [
{
"lang": "en",
"value": "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."
}
],
"metrics": [
{
"cvssV4_0": {
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 7.1,
"baseSeverity": "HIGH",
"privilegesRequired": "NONE",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "NONE",
"userInteraction": "PASSIVE",
"vectorString": "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",
"version": "4.0",
"vulnAvailabilityImpact": "HIGH",
"vulnConfidentialityImpact": "NONE",
"vulnIntegrityImpact": "NONE"
},
"format": "CVSS"
},
{
"cvssV3_1": {
"attackComplexity": "LOW",
"attackVector": "NETWORK",
"availabilityImpact": "HIGH",
"baseScore": 6.5,
"baseSeverity": "MEDIUM",
"confidentialityImpact": "NONE",
"integrityImpact": "NONE",
"privilegesRequired": "NONE",
"scope": "UNCHANGED",
"userInteraction": "REQUIRED",
"vectorString": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H",
"version": "3.1"
},
"format": "CVSS"
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-787",
"description": "Out-of-bounds Write",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"dateUpdated": "2026-09-04T14:32:25.391Z",
"orgId": "83251b91-4cc7-4094-a5c7-464a1b83ea10",
"shortName": "VulnCheck"
},
"references": [
{
"name": "GitHub Issue #2094",
"tags": [
"issue-tracking"
],
"url": "https://github.com/huggingface/tokenizers/issues/2094"
},
{
"tags": [
"product"
],
"url": "https://github.com/huggingface/tokenizers"
},
{
"tags": [
"technical-description"
],
"url": "https://github.com/huggingface/tokenizers/blob/v0.23.2/tokenizers/src/models/bpe/model.rs"
},
{
"name": "VulnCheck Advisory: tokenizers BpeBuilder Buffer Overflow via merge token",
"tags": [
"third-party-advisory"
],
"url": "https://www.vulncheck.com/advisories/tokenizers-bpebuilder-buffer-overflow-via-merge-token"
}
],
"title": "tokenizers BpeBuilder Buffer Overflow via merge token",
"x_generator": {
"engine": "vulncheck-endgame"
}
}
},
"cveMetadata": {
"assignerOrgId": "83251b91-4cc7-4094-a5c7-464a1b83ea10",
"assignerShortName": "VulnCheck",
"cveId": "CVE-2026-85670",
"datePublished": "2026-09-04T14:32:25.391Z",
"dateReserved": "2026-09-04T13:32:36.617Z",
"dateUpdated": "2026-09-04T17:49:42.263Z",
"state": "PUBLISHED"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.2",
"vulnerability-lookup:meta": {
"epss": {
"cve": "CVE-2026-85670",
"date": "2026-09-16",
"epss": "0.0031",
"percentile": "0.23863"
},
"nvd": "{\"cve\":{\"id\":\"CVE-2026-85670\",\"sourceIdentifier\":\"disclosure@vulncheck.com\",\"published\":\"2026-09-04T15:17:44.993\",\"lastModified\":\"2026-09-04T18:18:06.417\",\"vulnStatus\":\"Received\",\"cveTags\":[],\"descriptions\":[{\"lang\":\"en\",\"value\":\"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.\"}],\"affected\":[{\"source\":\"disclosure@vulncheck.com\",\"affectedData\":[{\"vendor\":\"huggingface\",\"product\":\"tokenizers\",\"defaultStatus\":\"unaffected\",\"packageURL\":\"pkg:pypi/tokenizers\",\"versions\":[{\"version\":\"0\",\"lessThanOrEqual\":\"0.23.2\",\"versionType\":\"semver\",\"status\":\"affected\"}]}]}],\"metrics\":{\"cvssMetricV40\":[{\"source\":\"disclosure@vulncheck.com\",\"type\":\"Secondary\",\"cvssData\":{\"version\":\"4.0\",\"vectorString\":\"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\",\"baseScore\":7.1,\"baseSeverity\":\"HIGH\",\"attackVector\":\"NETWORK\",\"attackComplexity\":\"LOW\",\"attackRequirements\":\"NONE\",\"privilegesRequired\":\"NONE\",\"userInteraction\":\"PASSIVE\",\"vulnConfidentialityImpact\":\"NONE\",\"vulnIntegrityImpact\":\"NONE\",\"vulnAvailabilityImpact\":\"HIGH\",\"subConfidentialityImpact\":\"NONE\",\"subIntegrityImpact\":\"NONE\",\"subAvailabilityImpact\":\"NONE\",\"exploitMaturity\":\"NOT_DEFINED\",\"confidentialityRequirement\":\"NOT_DEFINED\",\"integrityRequirement\":\"NOT_DEFINED\",\"availabilityRequirement\":\"NOT_DEFINED\",\"modifiedAttackVector\":\"NOT_DEFINED\",\"modifiedAttackComplexity\":\"NOT_DEFINED\",\"modifiedAttackRequirements\":\"NOT_DEFINED\",\"modifiedPrivilegesRequired\":\"NOT_DEFINED\",\"modifiedUserInteraction\":\"NOT_DEFINED\",\"modifiedVulnConfidentialityImpact\":\"NOT_DEFINED\",\"modifiedVulnIntegrityImpact\":\"NOT_DEFINED\",\"modifiedVulnAvailabilityImpact\":\"NOT_DEFINED\",\"modifiedSubConfidentialityImpact\":\"NOT_DEFINED\",\"modifiedSubIntegrityImpact\":\"NOT_DEFINED\",\"modifiedSubAvailabilityImpact\":\"NOT_DEFINED\",\"Safety\":\"NOT_DEFINED\",\"Automatable\":\"NOT_DEFINED\",\"Recovery\":\"NOT_DEFINED\",\"valueDensity\":\"NOT_DEFINED\",\"vulnerabilityResponseEffort\":\"NOT_DEFINED\",\"providerUrgency\":\"NOT_DEFINED\"}}],\"cvssMetricV31\":[{\"source\":\"disclosure@vulncheck.com\",\"type\":\"Secondary\",\"cvssData\":{\"version\":\"3.1\",\"vectorString\":\"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H\",\"baseScore\":6.5,\"baseSeverity\":\"MEDIUM\",\"attackVector\":\"NETWORK\",\"attackComplexity\":\"LOW\",\"privilegesRequired\":\"NONE\",\"userInteraction\":\"REQUIRED\",\"scope\":\"UNCHANGED\",\"confidentialityImpact\":\"NONE\",\"integrityImpact\":\"NONE\",\"availabilityImpact\":\"HIGH\"},\"exploitabilityScore\":2.8,\"impactScore\":3.6}],\"ssvcV203\":[{\"source\":\"134c704f-9b21-4f2e-91b3-4a467353bcc0\",\"ssvcData\":{\"timestamp\":\"2026-09-04T17:49:21.636135Z\",\"id\":\"CVE-2026-85670\",\"options\":[{\"exploitation\":\"poc\"},{\"automatable\":\"no\"},{\"technicalImpact\":\"partial\"}],\"role\":\"CISA Coordinator\",\"version\":\"2.0.3\"}}]},\"weaknesses\":[{\"source\":\"disclosure@vulncheck.com\",\"type\":\"Secondary\",\"description\":[{\"lang\":\"en\",\"value\":\"CWE-787\"}]}],\"references\":[{\"url\":\"https://github.com/huggingface/tokenizers\",\"source\":\"disclosure@vulncheck.com\"},{\"url\":\"https://github.com/huggingface/tokenizers/blob/v0.23.2/tokenizers/src/models/bpe/model.rs\",\"source\":\"disclosure@vulncheck.com\"},{\"url\":\"https://github.com/huggingface/tokenizers/issues/2094\",\"source\":\"disclosure@vulncheck.com\"},{\"url\":\"https://www.vulncheck.com/advisories/tokenizers-bpebuilder-buffer-overflow-via-merge-token\",\"source\":\"disclosure@vulncheck.com\"},{\"url\":\"https://github.com/huggingface/tokenizers/issues/2094\",\"source\":\"134c704f-9b21-4f2e-91b3-4a467353bcc0\"}]}}",
"vulnrichment": {
"containers": "{\"adp\": [{\"title\": \"CISA ADP Vulnrichment\", \"metrics\": [{\"other\": {\"type\": \"ssvc\", \"content\": {\"id\": \"CVE-2026-85670\", \"role\": \"CISA Coordinator\", \"options\": [{\"Exploitation\": \"poc\"}, {\"Automatable\": \"no\"}, {\"Technical Impact\": \"partial\"}], \"version\": \"2.0.3\", \"timestamp\": \"2026-09-04T17:49:21.636135Z\"}}}], \"references\": [{\"url\": \"https://github.com/huggingface/tokenizers/issues/2094\", \"tags\": [\"exploit\"]}], \"providerMetadata\": {\"orgId\": \"134c704f-9b21-4f2e-91b3-4a467353bcc0\", \"shortName\": \"CISA-ADP\", \"dateUpdated\": \"2026-09-04T17:49:37.775Z\"}}], \"cna\": {\"title\": \"tokenizers BpeBuilder Buffer Overflow via merge token\", \"credits\": [{\"lang\": \"en\", \"type\": \"finder\", \"value\": \"George Chen\"}], \"metrics\": [{\"format\": \"CVSS\", \"cvssV4_0\": {\"version\": \"4.0\", \"baseScore\": 7.1, \"attackVector\": \"NETWORK\", \"baseSeverity\": \"HIGH\", \"vectorString\": \"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\", \"userInteraction\": \"PASSIVE\", \"attackComplexity\": \"LOW\", \"attackRequirements\": \"NONE\", \"privilegesRequired\": \"NONE\", \"subIntegrityImpact\": \"NONE\", \"vulnIntegrityImpact\": \"NONE\", \"subAvailabilityImpact\": \"NONE\", \"vulnAvailabilityImpact\": \"HIGH\", \"subConfidentialityImpact\": \"NONE\", \"vulnConfidentialityImpact\": \"NONE\"}}, {\"format\": \"CVSS\", \"cvssV3_1\": {\"scope\": \"UNCHANGED\", \"version\": \"3.1\", \"baseScore\": 6.5, \"attackVector\": \"NETWORK\", \"baseSeverity\": \"MEDIUM\", \"vectorString\": \"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H\", \"integrityImpact\": \"NONE\", \"userInteraction\": \"REQUIRED\", \"attackComplexity\": \"LOW\", \"availabilityImpact\": \"HIGH\", \"privilegesRequired\": \"NONE\", \"confidentialityImpact\": \"NONE\"}}], \"affected\": [{\"vendor\": \"huggingface\", \"product\": \"tokenizers\", \"versions\": [{\"status\": \"affected\", \"version\": \"0\", \"versionType\": \"semver\", \"lessThanOrEqual\": \"0.23.2\"}], \"packageURL\": \"pkg:pypi/tokenizers\", \"defaultStatus\": \"unaffected\"}], \"datePublic\": \"2026-06-10T00:00:00.000Z\", \"references\": [{\"url\": \"https://github.com/huggingface/tokenizers/issues/2094\", \"name\": \"GitHub Issue #2094\", \"tags\": [\"issue-tracking\"]}, {\"url\": \"https://github.com/huggingface/tokenizers\", \"tags\": [\"product\"]}, {\"url\": \"https://github.com/huggingface/tokenizers/blob/v0.23.2/tokenizers/src/models/bpe/model.rs\", \"tags\": [\"technical-description\"]}, {\"url\": \"https://www.vulncheck.com/advisories/tokenizers-bpebuilder-buffer-overflow-via-merge-token\", \"name\": \"VulnCheck Advisory: tokenizers BpeBuilder Buffer Overflow via merge token\", \"tags\": [\"third-party-advisory\"]}], \"x_generator\": {\"engine\": \"vulncheck-endgame\"}, \"descriptions\": [{\"lang\": \"en\", \"value\": \"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.\"}], \"problemTypes\": [{\"descriptions\": [{\"lang\": \"en\", \"type\": \"CWE\", \"cweId\": \"CWE-787\", \"description\": \"Out-of-bounds Write\"}]}], \"providerMetadata\": {\"orgId\": \"83251b91-4cc7-4094-a5c7-464a1b83ea10\", \"shortName\": \"VulnCheck\", \"dateUpdated\": \"2026-09-04T14:32:25.391Z\"}}}",
"cveMetadata": "{\"cveId\": \"CVE-2026-85670\", \"state\": \"PUBLISHED\", \"dateUpdated\": \"2026-09-04T17:49:42.263Z\", \"dateReserved\": \"2026-09-04T13:32:36.617Z\", \"assignerOrgId\": \"83251b91-4cc7-4094-a5c7-464a1b83ea10\", \"datePublished\": \"2026-09-04T14:32:25.391Z\", \"assignerShortName\": \"VulnCheck\"}",
"dataType": "CVE_RECORD",
"dataVersion": "5.2"
}
}
}
Loading…
Loading…
Experimental. This forecast is provided for visualization only and may change without notice. Do not use it for operational decisions.
Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Loading…
The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the
vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base
model, served locally by ML-Gateway.
They have not been verified by an analyst and are provided for guidance only.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Loading…
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.
Loading…