CVE-2026-73555 vllm-project CVE debrief

Who should care

Defenders managing vLLM deployments, especially those exposed to unauthenticated requests, should assess their exposure and prioritize updating to version 0.26.0 or later. This includes operators, platform administrators, vulnerability management teams, and security teams who need to verify exposure, assess potential impact, and apply necessary updates or mitigations.

Why it matters

CVE-2026-73555 is an information disclosure vulnerability in vLLM that allows unauthenticated attackers to access sensitive information. Defenders should prioritize verifying exposure, especially in deployments with unauthenticated access to vLLM endpoints, and update to version 0.26.0 or later.

• Potential disclosure of sensitive information such as OS usernames and paths
• Possible exposure of internal package structure and line numbers
• Risk of unauthorized access to endpoint handler names
• Need for verification of vLLM deployment exposure and remediation

Technical summary

The vLLM inference and serving engine for large language models, prior to version 0.26.0, contains a vulnerability that allows unauthenticated attackers to disclose sensitive information such as OS username, home and virtual-environment paths, Python version, internal package structure, line numbers, and endpoint handler names through malformed JSON requests to certain endpoints. This issue is fixed in version 0.26.0. Defenders should prioritize verifying exposure to this information disclosure vulnerability, especially those managing vLLM deployments.

Defensive priority

Defenders should prioritize verifying exposure to this information disclosure vulnerability, especially those managing vLLM deployments.

Recommended defensive actions

• Verify vLLM deployment exposure to unauthenticated JSON requests
• Assess the impact of potential information disclosure
• Update to version 0.26.0 or later if vulnerable
• Monitor for suspicious requests to /v1/chat/completions, /v1/completions, /tokenize, and /detokenize
• Review compensating controls for exposed systems while remediation is scheduled and verified
• Check relevant monitoring, detection, and logs for exposed assets that need extra review
• Track exceptions, retest remediated assets, and close the item only after evidence is documented

Evidence notes

The CVE record and NVD entry provide details on the vulnerability in vLLM, an inference and serving engine for large language models, which could allow unauthenticated disclosure of sensitive information via malformed JSON requests.

Sources and references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.