Critical "Bleeding Llama" Vulnerability Puts 300,000 Ollama Deployments at Risk

A critical vulnerability, dubbed "Bleeding Llama," has been discovered in Ollama, a platform used for running large language models locally. This heap out-of-bounds read vulnerability could potentially expose up to 300,000 Ollama deployments to information theft. The vulnerability can be exploited remotely and does not require any authentication, making it a significant security risk for users.

The Bleeding Llama vulnerability allows attackers to read arbitrary data from the memory of an Ollama instance, which could include sensitive information or model parameters. Given the widespread use of Ollama for developing and deploying AI models, the potential impact of this vulnerability is substantial.

Ollama developers have been notified of the issue and are expected to release a patch to address the vulnerability. Users are strongly advised to update their Ollama installations as soon as a fix becomes available to mitigate the risk of information theft and unauthorized access.