From the course: Agentic AI Fundamentals: Architectures, Frameworks, and Applications
Join today to access over 24,500 courses taught by industry experts.
Security challenges
From the course: Agentic AI Fundamentals: Architectures, Frameworks, and Applications
Security challenges
- Deploying systems that humans don't monitor, or control has both benefits and challenges when it comes to security. Primarily, these challenges can be categorized into four important concepts. Let's review them together. First, AI agents must handle many types of user inputs. When they involve multiple steps, these inputs can be unpredictable. For example, an AI agent assists a customer in finding and purchasing a product on a website. However, a customer decides to split a purchase between several credit cards and PayPal. This practice is not unheard of. However, for this specific agent, there were no security approaches to deal with this unpredictable process. As a result, personal data was exposed, but not because there was any malicious intent. The problem was that developers had not considered the sequence of processing credit cards, so their AI agent didn't know what to do. The developers could argue they did not foresee this occurring, thus did not create a security system to…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
