"Unpatched Critical Vulnerabilities Open AI Models to Takeover"

Researchers discovered nearly a dozen critical vulnerabilities in the infrastructure used by Artificial Intelligence (AI) models, along with three high- and two medium-severity bugs, which could put companies at risk as they rush to capitalize on AI. The affected platforms host, deploy, and share Large Language Models (LLMs), as well as other Machine Learning (ML) platforms and AIs. They include Ray, MLflow, ModelDB, and H20 version 3. Protect AI, an ML security company, revealed the findings on November 16 as part of its AI-specific bug bounty program. It notified software maintainers and vendors about the vulnerabilities, giving them 45 days to fix the problems. This article continues to discuss the security holes in the infrastructure used by AI models.

Dark Reading reports "Unpatched Critical Vulnerabilities Open AI Models to Takeover"

Submitted by grigby1