The vulnerability of the library for working with large language models NVIDIA TensorRT-LLM, related to deficiencies in the deserialization mechanism, allows attackers to gain access to the TRTLLM server.

The vulnerability of the library for working with large language models like NVIDIA TensorRT-LM is related to deficiencies in the deserialization mechanism. Exploiting this vulnerability could allow attackers to gain access to the TRTLLM server...