Shadow AI: The Cybersecurity Threat Companies Can’t See | Interview with Timo Bozsolik-Torres

AI is rapidly being adopted across companies worldwide, but many security leaders don’t actually know where it’s being used. This growing blind spot is known as “Shadow AI.” In this video, Timo Bozsolik-Torres, Head of AI for Cybersecurity at SandboxAQ, explains how AI tools, models, agents, and MCP servers are quietly appearing across enterprise environments, and why that creates major security risks. With roughly 80% of companies already using AI, organizations must now answer critical questions: -What AI models are running inside our systems? -Are those models safe and secure? -What happens when AI agents get access to real company infrastructure? From prompt injection and toxic outputs to autonomous agents accidentally deleting production databases, AI introduces entirely new attack surfaces. Because models are probabilistic and constantly evolving, traditional cybersecurity playbooks don’t always apply. To address this challenge, new approaches are needed: -Scanning AI models for safety and security vulnerabilities -Discovering which models and agents are running inside enterprise codebases -Building an inventory of AI systems in use across an organization -Researching new attack vectors and defenses for emerging AI technologies As AI becomes more integrated into business operations, securing these systems is critical to preventing costly failures and malicious attacks. Subscribe for more insights on AI, cybersecurity, and emerging technologies. #sandboxaq #lqms #aqtiveguard