THREAT OPS › Threat News › [NVD] CVE-2026-73558 (MEDIUM 5.3) — vLLM is an inference and serving engine for large language models. Prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to
[NVD] CVE-2026-73558 (MEDIUM 5.3) — vLLM is an inference and serving engine for large language models. Prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to
CVE-2026-73558 CVSS: 5.3 MEDIUM Published: 2026-08-13T15:20:18.220
vLLM is an inference and serving engine for large language models. Prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to receive a partial or complete copy of another user'
Indicators of compromise
- CVE-2026-73558cve
Original source: https://nvd.nist.gov/vuln/detail/CVE-2026-73558