Agentic Risk

Agentic risk refers to the new category of exposure created when AI systems are given the ability to take actions on their own, such as browsing the web, sending emails, executing code, or making purchases on a user's behalf. Unlike a chatbot that only produces text for a person to review, an AI agent can carry out multi-step tasks that change systems and data, so a single mistake or manipulation can have direct financial, legal, or operational consequences. For compliance and risk teams, this shifts the focus from what an AI says to what it is permitted to do, making access controls, action logging, and human approval steps central parts of governance.

1 item

ResearchGlobal2026-10-09

GWU Formula Predicts AI Chatbot Behavioral Failure Before It Happens

Researchers at George Washington University have developed a mathematical formula to predict when AI chatbots will produce unsafe or undesirable outputs. The formula targets on-device AI models that operate without cloud-based safety filters or live monitoring. The researchers validated their approach across seven open-weight AI models and propose an in-model early-warning system as a practical control.

AI safety on-device AI open-weight models behavioral monitoring agentic risk model governance George Washington University