Nvidia has released a software update to address 'critical' vulnerabilities in its Triton inference server, a tool designed to enhance the deployment of AI models. These vulnerabilities, identified by cybersecurity firm Wiz, could allow unauthorized control over AI models, data theft, and manipulation of responses if not mitigated. The vulnerability chain begins with a minor bug that leaks sensitive internal data, which attackers could exploit to gain access to server components and escalate privileges. Although the precise number of users for Triton is unclear, significant companies like Microsoft and Amazon are among those utilizing the platform. Nvidia has not confirmed any instances of these vulnerabilities being exploited but urges all users to update to the patched version (25.07 or newer) to safeguard their systems. This incident highlights ongoing security challenges as advanced technologies evolve, potentially affecting various sectors including crypto, where vulnerabilities have led to significant financial losses.

Source 🔗